# Citegram > Citegram asks your buyers' questions in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, and shows whether your brand is mentioned, where it ranks, and which sources the AI relied on. --- # AI Visibility for Local Businesses: Get Recommended by ChatGPT Source: https://www.citegram.com/blog/ai-visibility-local-business/ Updated: 2026-10-10 Local business ChatGPT recommendations, and AI answers in general, largely depend on what the sources an assistant reads say about you: your Google Business Profile, review platforms, directories, local press and your own website. If those sources describe your service, location and reputation clearly and consistently, you give an assistant something solid to recommend. If they're thin or contradictory, a competitor with a better-documented presence is the safer pick. People now ask assistants "best dentist in Lyon" or "plumber near me open on Sunday" and get a short list with reasons. This guide covers what local prompts look like, where assistants find local information, a step-by-step plan for AI SEO for local business and the honest limits of measuring it. ## Key takeaways - AI visibility for a local business is mostly local SEO done well: a complete Business Profile, real reviews and a website that states what you do and where. - Google says Business Profiles can help your products and services be visible in its AI responses as well as in regular search results. - Google's local ranking depends on relevance, distance and prominence, and prominence includes how many reviews you have and how many websites link to you. - Local answers depend on where the person asking is, so "near me" prompts give different results in different places. - To measure local AI visibility, write the city into your prompts and run the same set repeatedly, rather than trusting one check. ## What local prompts look like Local questions to an assistant tend to fall into a few patterns: - **Category plus city:** "best dentist in Lyon", "accountant in Bordeaux for freelancers". - **Near me:** "plumber near me", "emergency locksmith near me open now". - **Need plus constraint:** "dentist in Lyon who takes new patients and speaks English", "plumber in Lille who can come today for a leak". The constraints matter. An assistant can only match "open on Sunday" or "speaks English" to your business if those details are written somewhere it reads. ### Why "near me" is different Google explains that it estimates your location when you search from several signals: [device location, saved home and work addresses, past activity and IP address](https://support.google.com/websearch/answer/179386). The same "near me" query can return different businesses depending on where the person is. Other assistants may also adapt local answers to location, depending on their settings. ## Where AI assistants get local information ### Google AI Overviews Google is the most explicit. Its guide to [optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) states that Merchant Center and Google Business Profiles "can help your products and services to be visible in both AI responses and other Google Search results". Its best practices also include keeping Business Profile data current. Google's help page on [improving your local ranking](https://support.google.com/business/answer/7091) lists three factors for local results: - **Relevance:** how well a Business Profile matches what someone is searching for. - **Distance:** how far the business is from the person searching. - **Prominence:** how well known the business is, based partly on "how many websites link to your business and how many reviews you have". Google also says there's "no way to request or pay for a better local ranking". ### ChatGPT, Gemini, Perplexity and others Other assistants don't document how they source local answers in the same detail. When they search the web, the sources cited for local questions commonly include review platforms, directories, map listings, local media and business websites. Which ones dominate varies by assistant, city and trade, so check the citations for your own prompts rather than assuming. Our guide on [how AI chooses sources](/blog/how-ai-assistants-choose-sources) explains the general mechanics. | Source | What it gives an assistant | What to check | |---|---|---| | Google Business Profile | Category, address, hours, attributes, reviews | Complete, verified, current | | Review platforms | Independent opinions and ratings | Recent reviews, replies, accurate details | | Directories and trade listings | Confirmation of name, address, services | Same details as your profile | | Local press and partners | Proof that you're known locally | Accurate mentions and links | | Your website | Services, areas, prices, credentials | Facts in text, easy to find | ## How to get recommended by AI as a local business Nothing guarantees a recommendation. These steps follow Google's published guidance and give every assistant clearer material to work with. 1. **Complete and verify your Business Profile.** Google says businesses with complete and accurate information are more likely to show up in local results. Fill in the category, address, hours, special hours and attributes, add photos and verify the profile. 2. **Earn reviews and reply to them.** Google says more reviews and positive ratings can help your local ranking, and that replying shows you value feedback. Ask real customers at the right moment. Never buy reviews or write fake ones. 3. **State the facts on your website.** One clear page per real service, the areas you actually serve, opening hours and prices or price ranges where you can give them, all in text. An assistant can only repeat "open Saturdays" or "treats children" if it's written somewhere. 4. **Keep details consistent everywhere.** Your name, address, phone number, hours and services should match across your profile, website, directories and review platforms. Contradictions make you harder to recommend with confidence. 5. **Get known locally.** Local press, trade associations, suppliers and community sites that mention and link to you feed prominence. Google warns that seeking inauthentic mentions "isn't as helpful as it might seem", so earn them. 6. **Don't mass-produce city pages.** Thin, near-identical pages for towns you don't really serve help no one. Our guide to [getting cited through review sites](/blog/get-cited-by-ai-review-sites) goes deeper into steps 2 and 5. For Google specifically, see [how to get cited in Google AI Overviews](/blog/get-cited-in-google-ai-overviews). ## The limits of measuring local AI visibility - **Answers depend on location.** A "near me" prompt answered in central Lyon may not match one answered across town. - **Answers depend on the account.** Personalization, memory and the signed-in account can shape the response. - **Answers vary between runs.** One check is an anecdote. A pattern over repeated runs is evidence. The practical fix is to write the city or neighbourhood into your tracked prompts ("best dentist in Lyon 6e") instead of relying on "near me". Our method for [checking if ChatGPT recommends your brand](/blog/how-to-check-if-chatgpt-recommends-your-brand) applies to local businesses too. ## How to track local AI recommendations with Citegram Citegram is a Chrome extension that asks your prompts in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews and records the answers. 1. **Write your local prompts.** Use the patterns above, with the city written in. The free [AI prompt generator](/tools/ai-prompt-generator) can help you draft variations. 2. **Run them across assistants.** Citegram asks each prompt in background tabs, using the accounts already signed in to your browser, with no API key. Questions count toward your own assistant plans. 3. **See who gets recommended.** Each answer records whether your business is mentioned, its position and the sources cited. For Google, it records whether an AI Overview appeared and which sources it cited. 4. **Find the sources that matter.** Reports show the most cited domains, so you know which review platforms or directories to prioritize. 5. **Compare and re-run.** Share of voice per brand and per assistant shows how often nearby competitors are named, and trends over repeated runs show whether that moves. First results arrive in about 10 minutes. One honest caveat: Citegram runs from your browser, so answers reflect your accounts and your location. It doesn't simulate a customer elsewhere, and it doesn't track Google Maps or the local pack. Brands with shops and products, rather than a single location, may prefer our page on [AI visibility for e-commerce brands](/for/ecommerce). ## FAQ ### How do I get my business recommended by ChatGPT? Make sure the sources an assistant is likely to read describe you clearly: a complete Business Profile, recent genuine reviews, consistent directory listings and a website that states your services, areas and hours in text. Then check what ChatGPT actually says for prompts with your city in them, and repeat over time. ### Does Google Business Profile affect AI Overviews? Google says Merchant Center and Google Business Profiles can help your products and services be visible in its AI responses as well as in regular search results. Keeping your profile complete and current is one of the clearest steps you can take. ### Why does ChatGPT recommend a competitor instead of my business? Usually because the sources it read describe the competitor better: more reviews, more complete listings or clearer details matching the prompt. Check which sources were cited and compare what they say about each business. ### Can I pay to be recommended by AI assistants? Not in their organic answers. Google states there's no way to request or pay for a better local ranking, and organic recommendations in assistants are separate from any ads. The reliable route is a well-documented, well-reviewed presence. ## See whether assistants recommend you If you want to know whether AI assistants recommend your business for the questions local customers ask, start with the free plan: 1 prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # Do Backlinks Matter for AI Search? Brand Mentions vs Links Source: https://www.citegram.com/blog/backlinks-vs-brand-mentions-ai/ Updated: 2026-10-10 Yes, backlinks still matter for AI search, but mostly indirectly. Links help search engines discover and rank your pages, and assistants that search the web before answering tend to read pages that rank. Whether an assistant actually names your brand, and how it describes you, depends more on brand mentions: how often, where and in what words the web talks about you, with or without a link. So the useful question isn't "links or mentions?" but what each one does. This article separates the two jobs, looks at what one large correlation study found, and gives a practical plan for B2B teams that already invest in link building and wonder where AI visibility fits. ## Key takeaways - Backlinks still help in AI search, because they help your pages get discovered and rank in the searches that web-connected assistants run. - Brand mentions, linked or not, shape whether an assistant connects your brand with a category and how it describes you. - An Ahrefs study of 75,000 brands found that branded web mentions correlated more strongly with Google AI Overview visibility than backlinks did, though correlation does not prove cause. - Bought links and manufactured mentions carry risk, and Google says inauthentic mentions are less helpful than they seem. - The practical test is to check which domains assistants cite for your prompts and whether they name you. ## Two jobs: being found and being named An AI answer to "best CRM for a 20-person startup" looks like one paragraph, but it usually involves two separate steps. **Being found.** Assistants with web search run one or more queries, read some of the pages that come back and cite a few. Your page can only be read if it is crawlable, indexed and ranks for those searches. Our guide to [query fan-out](/blog/query-fan-out-explained) shows how one prompt can turn into several searches. **Being named.** The answer is a synthesis. Which brands make the shortlist, and the adjectives attached to them, come from what the sources say, or from what the model learned in training when it doesn't search. A roundup that recommends HubSpot and Pipedrive but not you can keep you off the shortlist even if your own page ranks. Links mostly help with the first job. Mentions mostly help with the second. Most brands need both. ## Why backlinks still matter for AI search ### Links help pages get discovered and ranked Search engines still find many new pages by following links, and links remain part of how pages earn authority. A comparison page with no links pointing to it is harder to rank for "HubSpot vs Pipedrive", and a page that doesn't rank for the searches an assistant runs is unlikely to be read. ### Access comes before authority Links can't help a page that assistants can't reach. OpenAI's [crawler documentation](https://developers.openai.com/docs/bots) says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Check robots.txt and bot protection first, as covered in [how to rank in ChatGPT](/blog/how-to-rank-in-chatgpt). ### A link usually comes with a mention A link from a relevant article rarely stands alone. It sits in a sentence that names your brand and says something about it. That text is what a model reads, so a good link is often a good mention too. ## Why brand mentions matter more than they used to ### Mentions build category association To recommend you for "CRM for startups", an assistant needs sources that connect your brand with that category and use case. Repeated, independent mentions in context build that connection. As far as the text goes, an unlinked mention in a comparison article says the same thing as a linked one. ### The description travels with the mention Assistants paraphrase. If many sources describe Pipedrive as a simple pipeline tool for small sales teams, that framing is likely to reach answers. "Easy to set up", "expensive" and "best for enterprises" are all mentions, and only some of them help you. Links alone carry no such description, apart from anchor text. ### Third-party pages are often what gets cited For comparison and "best X" prompts, assistants frequently cite review sites, roundups and forum threads rather than brand websites. What those pages say about you shapes the answer. We cover how to earn honest presence there in [how to get cited by AI on review sites, Reddit and comparisons](/blog/get-cited-by-ai-review-sites). ## What the research says, and what it doesn't In May 2025, Ahrefs published [an analysis of 75,000 brands](https://ahrefs.com/blog/ai-overview-brand-correlation/) that compared several factors with how often each brand was mentioned in Google AI Overviews, using Spearman correlations: | Factor | Correlation with AI Overview mentions | |---|---| | Branded web mentions | 0.664 | | Branded anchors | 0.527 | | Branded search volume | 0.392 | | Domain Rating | 0.326 | | Referring domains | 0.295 | | Backlinks | 0.218 | Read it with care: - **It is correlation.** The authors say so. Big brands get more mentions, links and AI visibility at the same time. - **It covers one surface.** Google AI Overviews, not ChatGPT, Perplexity or Claude. - **The sample skews large.** Only domains with a Domain Rating above 40 were included. Google's own [guide to generative AI features in Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) points in the same direction from a different angle. It says these features "can show what's being said about products and services across the web, including in blogs, videos, and forum discussions", and also that "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." The fair reading: how widely and how credibly the web talks about you seems to matter a lot for AI answers, links still matter for search, and nobody has shown that either one can be gamed. ## Backlinks vs brand mentions at a glance | | Backlinks | Brand mentions (linked or not) | |---|---|---| | Main job in AI search | Help pages get discovered and rank for the searches assistants run | Tell assistants your brand belongs in a category, and how to describe it | | Where it matters most | Assistants and AI Overviews that search the web | Live search answers and, over time, what models learn | | Carries a description | Only through anchor text and nearby copy | Yes, the sentence around your name | | Risk when faked | Link spam penalties | Wasted effort and reputational damage | ## A practical plan: where to put the effort 1. **Keep the foundations.** Make sure search and AI crawlers can reach your key pages, and keep earning links to the pages that need to rank: comparison, pricing and use-case pages. 2. **Find the pages that shape answers.** List the domains assistants cite for your buyer prompts. Those are your mention targets, whether or not they would ever link to you. 3. **Judge outreach by mention quality.** A clear, accurate mention in a roundup that assistants cite can be worth more for AI answers than a followed link on an unrelated blog. Ask: is this page read for our prompts, and does it describe us correctly? 4. **Write the sentence you want repeated.** Give reviewers, journalists and partners a factual one-line description: who the product is for, what it does, what it costs. Consistent wording across sources makes a consistent answer more likely. 5. **Correct what's wrong.** Outdated pricing or a retired feature on a widely read page can end up in answers. Contact the author with evidence. 6. **Don't buy either.** Paid links break search engine rules, and planted or fake mentions tend to be spotted by readers, moderators and spam systems. ## Measure mentions and citations with Citegram Backlink tools tell you who links to you. They don't tell you whether assistants name you. Citegram is a Chrome extension that covers the answer side (it does not track backlinks): 1. **Add unbranded buyer prompts.** Questions like "best CRM for a 20-person startup". The [AI prompt generator](/tools/ai-prompt-generator) can give you a starting list. 2. **Run them in the real assistants.** Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, using the accounts already signed in to your browser. There's no API key, and questions count toward your own plan usage. 3. **See whether you're named and where.** Each answer records whether your brand is mentioned, its position, and the sources and links it cited. 4. **Compare cited domains with your link profile.** The report of most cited domains shows which sites shape answers in your category. Domains that are cited often but never mention you are your outreach list. 5. **Re-run after each PR push.** Share of voice per brand and per assistant, tracked over repeated runs, shows whether new mentions are reaching the answers. Answers come from your own accounts, so personalization and location can affect them. Judge trends, not single runs. SaaS teams can see more on the [SaaS page](/for/saas). ## FAQ ### Do backlinks help you get cited by ChatGPT? Indirectly. When ChatGPT searches the web, it reads pages its crawler can reach and that rank for the searches it runs, and links help pages rank. Whether it recommends you also depends on what those sources say about you. ### Do unlinked brand mentions count for AI search? They can. An assistant reads text, so a mention without a link still tells it that your brand exists and how others describe it. An Ahrefs correlation study found branded web mentions were more closely associated with Google AI Overview visibility than backlinks. ### Should I stop building links for GEO? No. Links still support the rankings that web-connected assistants depend on. What changes is how you pick targets: favor pages that are cited for your buyer prompts and describe you accurately, not only pages with strong link metrics. ### How do I find where my brand should be mentioned? Run your buyer prompts in the assistants your audience uses and list the domains they cite. Then check which of those pages mention competitors and not you. Our guide to [how AI assistants choose sources](/blog/how-ai-assistants-choose-sources) explains what tends to get a page selected. ## See which sources shape your answers Before you plan the next link or PR campaign, check which domains assistants cite for your category and whether they name you. The free plan tracks 1 prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # Claude SEO: How Claude Finds and Cites Sources Source: https://www.citegram.com/blog/claude-seo/ Updated: 2026-10-10 Claude SEO is the work of making your pages easy for Claude to find when it searches the web, and easy to quote when it cites its sources. Anthropic documents that Claude can search the live web to ground its answers, that every response using web search includes citations, and that it runs three separate bots. Getting cited by Claude starts with letting the right bots in, then publishing pages that answer the searches Claude runs. When a buyer asks Claude "what's the best CRM for a 20-person startup?", you want to be in the answer and ideally in the sources. This guide covers only what Anthropic documents about Claude's search and crawlers, what that means for your site, and how to measure whether Claude cites you. ## Key takeaways - Claude answers either from its training data or by searching the web. When it searches, Anthropic says every response includes citations to its sources. - Anthropic runs three bots: ClaudeBot for training, Claude-SearchBot for search indexing and Claude-User for fetches triggered by user questions. - Blocking Claude-SearchBot or Claude-User may reduce your visibility in Claude's search results. Blocking ClaudeBot only signals that your future content should be excluded from training. - Anthropic says its bots honor robots.txt and advises against blocking them by IP address, because the bot then can't read your robots.txt. - Anthropic doesn't publish ranking factors for Claude's search, so the reliable method is to measure which sources Claude cites for your buyer prompts. ## How Claude answers: training data or live web search Claude can answer a question from what it learned during training, or by searching the web first. Anthropic's help article on [using web search in Claude](https://support.claude.com/en/articles/10684626-enable-and-use-web-search) explains that for topics that benefit from current information, "Claude invokes a search tool to inform and ground its generated responses with content from the live web". It adds that "every response includes citations", along with source links and, when appropriate, relevant quotes. Depending on the plan and interface, web search is a setting the user turns on, or Claude simply searches when it helps. On Team and Enterprise plans, an owner has to enable it for the workspace first. ### When Claude decides to search Anthropic's documentation for the [web search tool in the Claude API](https://platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool) lists the requests that make Claude search. Two of them describe most B2B buying questions: - **Current prices**, rates or statistics. - **Information about specific organizations, people or products** that might have changed. A prompt like "which CRM is cheapest for a 20-person team" fits both. Questions about stable knowledge, such as definitions or coding concepts, are usually answered without searching. This list comes from the API documentation, and Anthropic doesn't say the Claude apps follow exactly the same rules, but it is the clearest official description of when Claude searches. ### One prompt, several searches The same API page says that "simple factual queries typically use 1–3 searches; comparative or multientity research can use 10 or more". A shortlist question that names several vendors is exactly that kind of research. Your page can be read because it answers one of those narrower searches, even if it doesn't match the user's original wording. We explain this pattern in [query fan-out explained](/blog/query-fan-out-explained). ## Anthropic's three bots and what blocking each one does Anthropic's [crawler help article](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) describes three bots, each with its own robots.txt token. | Bot | What Anthropic says it does | What happens if you block it | |---|---|---| | `ClaudeBot` | Collects web content that could contribute to model training | Signals that your future materials should be excluded from training datasets | | `Claude-SearchBot` | Navigates the web to improve the relevance and accuracy of search responses | Prevents indexing for search, which may reduce your visibility and accuracy in search results | | `Claude-User` | Accesses websites when individuals ask Claude questions | Prevents retrieval of your content in response to a user query, which may reduce your visibility | For Claude SEO, the two that matter most are Claude-SearchBot and Claude-User. ClaudeBot is a separate training decision, and you can block it while keeping the other two allowed. Anthropic also says its bots honor robots.txt, respect anti-circumvention technologies such as CAPTCHAs, and support the non-standard `Crawl-delay` directive. Each subdomain needs its own robots.txt rules. For a side-by-side view with OpenAI, Perplexity and Google, see our guide to [AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt). ## What Anthropic doesn't document Be careful with any guide that claims to know Claude's "ranking factors". Anthropic's public documentation leaves several things open. | Documented | Not documented | |---|---| | Claude searches the web and cites sources in search responses | How results are ranked or selected | | Three bots and the effect of blocking each one | Which search index provides text results | | Search results in the API carry a page age (when the site was last updated) | How often a page is recrawled | | Claude may use your location, inferred from your IP address, for localized results | How many pages Claude reads per answer | One detail is worth knowing: Anthropic's help center says image results in Claude are powered by Bing. It doesn't name a provider for text search, so don't assume that ranking in Bing is the same as being found by Claude. ## What this means for your site Nothing on this list guarantees a citation. It removes the obstacles Anthropic documents and makes your pages easier to use. 1. **Allow Claude-SearchBot and Claude-User.** Check that robots.txt doesn't disallow them, including through a broad `User-agent: *` rule. The [AI robots.txt generator](/tools/ai-robots-txt-generator) can help you write separate rules for training and search bots. 2. **Check your CDN and firewall.** A "block AI bots" setting can stop Claude's bots before robots.txt is read. Anthropic warns that IP blocking may not work reliably as an opt-out, because the bot can't read your robots.txt. 3. **Make passages easy to quote.** In the API, each web search citation includes up to 150 characters of cited text. Short, self-contained sentences that state a fact (for example, "Plans start at $15 per user per month") are easier to attribute than long marketing paragraphs. 4. **Keep pages visibly current.** Claude searches when information might have changed. Put an updated date on pricing, comparison and "best of" pages, and keep the facts on them accurate. 5. **Be present on third-party pages.** For "best X for Y" questions, assistants often cite review sites, roundups and community threads. Our guide to [getting cited through review sites](/blog/get-cited-by-ai-review-sites) covers how to show up there. These are the same principles we describe in [how AI assistants choose the sources they cite](/blog/how-ai-assistants-choose-sources). Clear, specific pages are easier for any search-based assistant to use, and Claude is no exception. ## How to measure whether Claude cites you Anthropic doesn't offer site owners a report on Claude citations, so you need your own baseline. 1. **Choose 10 to 30 buyer prompts** without your brand name: category, use-case and comparison questions. Our guide to [choosing prompts to track](/blog/choosing-prompts-to-track) explains how, and the [AI prompt generator](/tools/ai-prompt-generator) gives you a first list. 2. **Run each one several times.** Answers vary between runs, and Claude may adapt localized results to the location inferred from your IP address. 3. **Record mentions, position, competitors and cited sources.** Note which of your pages appear and which third-party pages Claude relies on. 4. **Change one thing, then re-run the same prompts** a few weeks later and compare the trend. ## Track your Claude visibility with Citegram Citegram is a free Chrome extension with a dashboard that runs this process for you: 1. **Add your buyer prompts** in the dashboard. 2. **Run them in the real Claude app.** Citegram asks each prompt in a background tab with the Claude account already signed in to your browser. No API key, and questions count toward your own plan usage. 3. **See whether Claude named you and where.** Each run records whether your brand was mentioned and its position. 4. **Review the cited sources.** Citegram records the sources and links each answer cited, so you can see which pages Claude trusts for your prompts. 5. **Compare across assistants.** The same prompts run in ChatGPT, Gemini, Perplexity, DeepSeek, Grok and Google AI Overviews, with share of voice per brand and per assistant, the most cited domains and trends over repeated runs. First results arrive in about 10 minutes. For Google's assistant, see our guide to [Gemini SEO](/blog/gemini-seo). ## FAQ ### Does Claude search the web? Yes. Anthropic says Claude can search the web for topics that benefit from current information and ground its answer in live web content. Every response that uses web search includes citations, so you can see which pages it relied on. ### Should I block ClaudeBot? Block ClaudeBot if you don't want your future content used for Anthropic's model training. It is separate from Claude-SearchBot and Claude-User, so you can block it and still allow the bots that affect your visibility in Claude's search results. ### Does Claude use Google or Bing for web search? Anthropic's help center says image results are powered by Bing, but it doesn't name a provider for text web search. Claude-SearchBot also indexes content to improve Claude's search results. Treat any claim that Claude simply uses one search engine with caution. ### How do I get cited by Claude? Let Claude-SearchBot and Claude-User reach your pages, publish clear and current answers to the questions your buyers ask, and make sure the third-party pages Claude cites describe you accurately. Then measure which sources Claude cites for your prompts, since Anthropic doesn't publish how it selects them. ## See where you stand in Claude Find out whether Claude names and cites you today, before you change anything. The free plan tracks one prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # ChatGPT Says Wrong Things About Your Brand? How to Fix It Source: https://www.citegram.com/blog/fix-wrong-ai-information-about-brand/ Updated: 2026-10-10 If ChatGPT has wrong information about your company, you can't edit the model directly. What you can do is find the sources the answer relies on, correct the ones you control, ask the owners of the others to update theirs, flag the answer through official feedback tools, and check over time that the correction sticks. A wrong price, a retired feature or a competitor's product attributed to you can cost deals quietly, because buyers rarely tell you what an assistant said. This guide covers why assistants get brand facts wrong, how to trace an error to its source, how to correct AI information about your brand, and how to monitor the result. ## Key takeaways - No AI company lets a brand edit what its model says, so corrections have to happen in the sources the assistant reads. - Most wrong answers trace back to something specific: an outdated page on your site, a stale review profile, an old comparison article or a similarly named company. - When an assistant cites sources, they show you where to start. When it cites nothing, the error may come from training data and will change more slowly. - Fix your own site first, then the third-party pages assistants cite most for your prompts. - Feedback tools in ChatGPT, Gemini, Perplexity and Google AI Overviews reach the vendor, but none promises a specific correction or timeline. ## Why AI assistants get facts about your brand wrong An assistant answers from what the model learned in training and from pages it retrieves at question time. We explain the retrieval side in [how AI assistants choose the sources they cite](/blog/how-ai-assistants-choose-sources). Errors usually come from one of these: - **Outdated pages on your own site.** An old pricing page or a help article for a retired feature can still be indexed and read. - **Stale third-party profiles.** Review platforms and software directories often keep whatever was entered years ago. - **Old comparison articles.** They keep ranking long after their facts change. - **Name confusion.** If another company has a similar name, an assistant can blend the two. - **Missing information.** When no page states a fact clearly, an assistant may fill the gap with a plausible guess. ## Step 1: Diagnose where the wrong information comes from Don't rewrite pages until you know which page is the problem. 1. **Write down the exact prompt and the wrong claim.** "Does [your brand] have a free plan?" answered with "No" when you've had one for a year is a precise, checkable error. 2. **Run the prompt several times, in several assistants.** Answers vary between runs. A claim that appears every time points to a widespread source. A one-off may be noise. 3. **Open the cited sources.** Look for the wrong statement in each linked page. That page is usually the culprit. 4. **Look at the searches behind the answer.** ChatGPT often splits a question into several web searches, a behavior called [query fan-out](/blog/query-fan-out-explained). The pages ranking for those searches are the ones it reads. 5. **Spot training-data errors.** If nothing is cited and the claim matches how things were years ago, it likely comes from training. Fix current sources anyway, but expect slower change. ### Match the error to its likely source | What the assistant gets wrong | Where it often comes from | First fix | |---|---|---| | Old pricing or plans | Old pricing page, review profiles, roundups | Update your pricing page, then the profiles | | Features you dropped or never had | Old help docs, launch posts, comparisons | Update or redirect old pages, contact authors | | Wrong category or audience | Directory listings, vague homepage copy | State category and audience plainly | | Mixed up with another company | Similar names, missing company details | Publish clear company facts everywhere | | Wrong founding date or location | Old press, company databases | Update your about page and the databases | ## Step 2: Fix your own site first Your site is the source you fully control, and assistants that search the web can read it. - **Publish a clear facts page.** Your about page or a dedicated facts page should state in plain sentences what the product is, who it's for, current plans and prices, main integrations, and company details. Show a last-updated date. - **Write facts as text.** Prices inside images or behind a "contact sales" button leave the assistant to rely on what other sites say. - **Handle outdated pages.** Update old posts and help articles or redirect them. A note such as "This plan was retired" helps readers and assistants alike. - **Be consistent.** Use the same description, category and pricing wording across your homepage, docs and social profiles. - **Stay reachable.** OpenAI says sites that opt out of its search crawler [won't be shown in ChatGPT search answers](https://developers.openai.com/docs/bots). Our free [llms.txt generator](/tools/llms-txt-generator) can produce a summary file for AI agents, but treat it as a convenience, not a fix. As our [llms.txt guide](/blog/llms-txt) explains, Google says you don't need AI text files to appear in Google Search. ## Step 3: Fix the third-party sources Many brand answers cite pages you don't own. Start with the domains that appeared most often in your diagnosis. ### Review and listing profiles Claim your profiles on the review platforms and directories that show up in citations. Update descriptions, pricing and features, and reply to reviews that contain factual errors. Our guide on [getting cited through review sites](/blog/get-cited-by-ai-review-sites) covers how to do this within each platform's rules. ### Comparison articles and roundups Email the author with the specific error, the correct fact and a link that proves it. Keep it short. Publishers owe you nothing, so ask only for the correction. ### Wikipedia and Wikidata, only if you're notable Wikipedia isn't a listing service. Its [notability guideline for companies](https://en.wikipedia.org/wiki/Wikipedia:Notability_(organizations_and_companies)) requires significant coverage in multiple independent, reliable secondary sources. If an article about you exists and contains errors, the [conflict of interest guideline](https://en.wikipedia.org/wiki/Wikipedia:Conflict_of_interest) strongly discourages editing it yourself: disclose your affiliation and propose changes on the talk page, with sources. Wikidata has its own notability policy, so read it before creating an entry. ### Press and company databases Keep a press page with a current boilerplate so journalists copy the right facts, and ask outlets to correct published errors, with evidence. ## Step 4: Report the answer through official feedback channels Feedback won't rewrite an answer on demand, but it tells the vendor the answer is wrong. Use the routes each vendor offers: - **ChatGPT:** use the thumbs-down icon under the response and describe the problem. Check OpenAI's help center for its current reporting options. - **Gemini:** Google's page on [sending feedback in Gemini Apps](https://support.google.com/gemini/answer/13275746) describes a "Bad response" icon with an optional reason and comment, plus "Report legal issue" in the more menu. - **Perplexity:** its [help center](https://intercom.help/perplexity-ai/en/articles/10354888-need-support) says to open the three-dot menu below an incorrect response and select "Report". - **Google AI Overviews:** per [Google Search Help](https://support.google.com/websearch/answer/14901683), click thumbs down, choose "Report a problem" and a category, and add details. Be specific every time: the wrong claim, the correct fact and a public URL that proves it. If an answer may be defamatory, bring in your legal team. ## Step 5: Monitor whether the correction sticks Sources need to be recrawled and answers vary between runs, so one good answer proves little. Keep the prompts that produced the error plus a few related questions, rerun them on a schedule, and record whether the wrong claim still appears and which sources are cited. Our guide to [tracking your brand across AI assistants](/blog/track-brand-across-ai-assistants) explains how to set up the routine. ## Track corrections across assistants with Citegram Citegram is a free Chrome extension with a dashboard that handles the monitoring: 1. **Add the prompts that produced the error**, plus related buyer questions. 2. **Run them in the real assistants.** Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, using the accounts already signed in to your browser. No API key is needed, and questions count toward your own plan usage. 3. **Read each answer and its sources.** You see the answer text, whether and where your brand is mentioned, and the cited sources in rank order. For ChatGPT, you also see the web searches it ran. 4. **Find the pages to fix.** The most cited domains report shows which sites shape answers across your prompts. 5. **Rerun after each fix** and compare runs over time, assistant by assistant. Answers come from your own accounts, so memory, personalization and location can affect them. Citegram doesn't flag false claims for you: you read the answers and judge them. ## FAQ ### How do I correct wrong information ChatGPT says about my company? Check the links in its answers and the searches it runs to find the sources behind the error. Fix your own pages, ask third-party sites to update theirs, and flag the answer with the thumbs-down icon. Then rerun the same questions over the following weeks. ### Can I ask OpenAI to change what ChatGPT says about my brand? There's no documented way for a brand to edit ChatGPT's answers. Feedback on a response reaches OpenAI but doesn't guarantee a correction. Fixing the sources ChatGPT reads is the part you control. ### Why does ChatGPT still show old information after I updated my website? Your updated page may not have been recrawled yet, or ChatGPT may be citing another page, such as an old review profile. If it answers without searching, the claim may come from training data, which changes more slowly. ### How long does it take for AI answers to reflect a correction? There's no fixed timeline. Assistants that search the web can pick up corrected pages once recrawled, while claims learned in training can persist much longer. Judge progress over several weeks. ## See what assistants say about you Start with the prompt where the error shows up. The free plan tracks 1 prompt across all major assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # Gemini SEO: How to Get Your Brand Recommended by Gemini Source: https://www.citegram.com/blog/gemini-seo/ Updated: 2026-10-10 Gemini SEO is the work of making your brand easy for Google's Gemini app to find, trust and recommend when someone asks a question your product answers. Google documents that Gemini Apps can ground answers in content from the Google Search index, so most of Gemini SEO is Google SEO done well: indexable pages, clear facts and a strong presence on the pages Google already ranks. There are no published Gemini ranking factors and no way to buy a place in its answers. When a buyer asks Gemini "what's the best CRM for a 20-person startup?", you want to be on the shortlist it writes. This guide sticks to what Google actually documents about Gemini, explains how it relates to Google Search and AI Overviews, and gives a practical process to improve and measure your visibility. ## Key takeaways - Google documents grounding for Gemini Apps as "providing content from the Google Search index to the model at prompt time", so being indexed and well ranked in Google is the foundation of Gemini visibility. - Google publishes no list of Gemini-specific ranking factors, and its guidance for AI features says no special files, markup or schema are needed. - Google-Extended is the one Gemini-specific control: it governs training and grounding in Gemini Apps, but it doesn't affect inclusion or ranking in Google Search. - Gemini, AI Overviews and AI Mode are separate products, so being cited in one doesn't mean being cited in the others. - Gemini doesn't link sources on every answer, so measure mentions and positions over repeated runs, not just links. ## What Google documents about how Gemini answers Most claims about "the Gemini algorithm" are guesses. Here is what Google actually states, and what it leaves open. | What Google documents | What it doesn't document | |---|---| | Gemini Apps can be grounded with content from the Google Search index at prompt time | When the Gemini app decides to search for a given prompt | | Google-Extended manages training and grounding in Gemini Apps | How grounded pages are selected or ranked inside Gemini | | The Gemini API can generate one or several search queries per prompt | Whether the consumer app uses exactly the same process | | Gemini Apps sometimes show sources, and not every response includes them | How many pages Gemini reads per answer | ### Grounding in the Google Search index Google's [list of common crawlers](https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers) describes Google-Extended as covering "grounding (providing content from the Google Search index to the model at prompt time to improve factuality and relevancy) in Gemini Apps". That is the clearest official link between Gemini and Search: grounded content comes from the same index that powers Google results. A page Google hasn't indexed can't be part of it. ### What the Gemini API documentation shows For developers, Google documents [grounding with Google Search](https://ai.google.dev/gemini-api/docs/google-search) in the Gemini API step by step. The model "analyzes the prompt and determines if a Google Search can improve the answer", then "automatically generates one or multiple search queries and executes them", processes the results and returns a response with citations linking parts of the text to source URLs. This describes the API, not the Gemini app, and Google doesn't say the two work identically. It's still a useful model: one prompt can become several searches, and the pages that rank for them are the ones the model reads. ### Sources in the Gemini app Google's help article on [viewing related sources in Gemini Apps](https://support.google.com/gemini/answer/14143489?hl=en) says Gemini "sometimes" shows sources, and that "not all responses include related links or sources". So a Gemini answer can recommend your competitor by name without linking to anything. Brand mentions matter as much as citations here. ## Gemini, Google Search and AI Overviews All three draw on Google's search systems, but they are different products. | | Gemini app | AI Overviews | AI Mode | |---|---|---|---| | Where it appears | gemini.google.com and the Gemini mobile apps | Above classic results on a Google search | A conversational mode inside Google Search | | Links to sources | Sometimes, not on every answer | Supporting links on the overview | Supporting links in the answer | | Tracked by Citegram | Yes | Yes | No | For AI Overviews and AI Mode, Google's [AI features documentation](https://developers.google.com/search/docs/appearance/ai-features) is explicit: a page must be indexed and eligible to show with a snippet, and there are no extra technical requirements. Google says AI Overviews and AI Mode may use "query fan-out", issuing multiple related searches across subtopics. Our guides on [AI Overviews and your traffic](/blog/google-ai-overviews-traffic) and [Google AI Mode SEO](/blog/google-ai-mode) cover those two surfaces in detail. Google doesn't publish an equivalent eligibility page for the Gemini app, so treat the Search requirements as a minimum. ## Google-Extended: the one Gemini-specific control Google-Extended isn't a crawler: it's a robots.txt token Google reads as a permission. Google says it manages whether your content may be used to train future Gemini models and for grounding in Gemini Apps, and that it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal". That creates a real trade-off: - **Allow Google-Extended** if you want Gemini to be able to use your pages when it grounds answers. - **Disallow it** if keeping your content out of Gemini training matters more to you, knowing it also covers grounding in Gemini Apps and may limit how Gemini uses your content. - **Don't use it for AI Overviews or AI Mode.** Those follow Googlebot rules and snippet controls such as `nosnippet`. Our [AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt) reference compares it with other vendors' bots. ## How to improve your chances of being recommended by Gemini 1. **Get your key pages indexed and snippet-eligible.** Check product, pricing, comparison and use-case pages with URL Inspection in Search Console, and remove any accidental `nosnippet` or Googlebot block. 2. **Check your Google-Extended rule.** Open your robots.txt and confirm the token isn't disallowed by accident, for example by a template that blocks "all AI bots". 3. **Answer the searches behind the prompt.** A question like "best CRM for a 20-person startup with email automation" hides narrower searches: CRM pricing per user, email automation features, comparisons with HubSpot or Pipedrive. Cover those subtopics on the pages you already have. Google warns that creating pages for every variation "primarily to manipulate rankings or generative AI responses" breaks its scaled content abuse policy. See [query fan-out explained](/blog/query-fan-out-explained). 4. **State your facts plainly.** Who the product is for, what it costs, what it integrates with and how it compares. A clear sentence is easier to reuse than a vague tagline. 5. **Be present on the pages Google ranks.** For "best X for Y" questions, top results are often review sites, roundups and community threads. Make sure they list you with accurate, current information. This is classic SEO applied to the questions buyers ask assistants; see [SEO vs GEO](/blog/seo-vs-geo) for where the two differ. ## How to measure your Gemini visibility Google says Search Console counts traffic from AI features in Search, but its documentation doesn't describe any report for the Gemini app. You need your own baseline. 1. **Build a prompt set.** Pick 10 to 30 questions your buyers would ask, without your brand name in them. The [AI prompt generator](/tools/ai-prompt-generator) can give you a starting list. 2. **Run each prompt several times.** Answers vary between runs and accounts. 3. **Record four things:** whether you're mentioned, your position in the list, which competitors appear and which sources are linked, when there are any. 4. **Compare with AI Overviews** for the same questions, since the two can disagree. 5. **Watch referral traffic.** Our guide to [tracking AI traffic in GA4](/blog/track-ai-traffic-ga4) shows how to separate visits from AI assistants. ## Track Gemini and AI Overviews with Citegram Citegram is a free Chrome extension with a dashboard that automates this measurement: 1. **Add your buyer prompts** in the dashboard. 2. **Run them in the real Gemini app.** Citegram asks each prompt in a background tab with the Google account already signed in to your browser. No API key, and questions count toward your own plan usage. Answers come from your account, so personalization and location can influence them. 3. **See whether Gemini named you and where.** Each run records whether your brand was mentioned, its position and the sources the answer cited. 4. **Check AI Overviews for the same questions.** Citegram records whether an AI Overview appeared in Google and which sources it cited. It doesn't track AI Mode. 5. **Compare across assistants.** The same prompts run in ChatGPT, Perplexity, Claude, DeepSeek and Grok, with share of voice per brand and per assistant, the most cited domains and trends over repeated runs. First results arrive in about 10 minutes. For the Anthropic side, see our guide to [Claude SEO](/blog/claude-seo). ## FAQ ### Is Gemini SEO different from Google SEO? Mostly no. Google documents that Gemini Apps can be grounded with content from the Google Search index, and its guidance for AI features says the usual SEO best practices apply. The differences are measurement and the Google-Extended setting. ### Does Gemini use Google Search results? Google says grounding in Gemini Apps provides content from the Google Search index to the model at prompt time. It doesn't document when the app decides to search or how it picks pages, so a top Google ranking helps but doesn't guarantee a mention. ### Does blocking Google-Extended remove my site from Gemini? It limits how Google may use your content for Gemini training and for grounding in Gemini Apps, so it can reduce how Gemini uses your pages. It doesn't affect your inclusion or ranking in Google Search, and it doesn't control AI Overviews or AI Mode. ### How do I check if Gemini recommends my brand? Ask Gemini the questions your buyers ask, without your brand name, several times each, and note whether you appear, in which position and next to which competitors. A tool like Citegram runs those prompts in the Gemini app for you and tracks the results over time. ## See where you stand in Gemini Before you change anything, find out whether Gemini names you for the questions your buyers ask. The free plan tracks one prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # GEO for SaaS: How B2B Software Gets Recommended by AI Source: https://www.citegram.com/blog/geo-for-saas/ Updated: 2026-10-10 GEO for SaaS is the work of getting your software named when buyers ask AI assistants for a shortlist: "best CRM for a 20-person startup", "alternatives to HubSpot", "Pipedrive vs HubSpot". In practice it means clear comparison, pricing, docs and integration pages, an accurate presence on the review sites assistants cite, and share of voice measured against the competitors you meet in deals. Assistants are often asked to choose between software products. A buyer types a question with real constraints (team size, budget, integrations) and gets back three or four names. If you're not one of them, there's no lost-deal reason in your CRM: the deal never started. This guide covers the prompts that build SaaS shortlists, the pages that feed them, and how to measure AI visibility as a SaaS company. ## Key takeaways - SaaS shortlists form on a few prompt types: "best X for Y", "alternatives to [competitor]", "X vs Y", and use-case questions with integration or budget constraints. - Each prompt type needs a page that answers it directly, such as a fair comparison page, a public pricing page or a specific integration page. - Review sites and roundups often weigh as much as your own site in recommendation answers, so keep those profiles accurate. - Public docs and pricing give assistants checkable facts. Gated or image-only information leaves them to rely on other sites. - Measure share of voice against the competitors you lose deals to, per assistant, and change one thing at a time. If GEO is new to you, start with [what generative engine optimization is](/blog/what-is-generative-engine-optimization). ## Why GEO matters for SaaS - **Crowded categories.** Most SaaS categories have dozens of credible vendors, and buyers ask an assistant to narrow the list before visiting any website. - **Constraint-heavy questions.** "Help desk that integrates with Slack and Jira" is exactly the kind of long, specific question assistants are asked to answer. - **Fact-heavy decisions.** Price per seat, free plans, integrations and security certifications are facts that assistants can extract, compare and get wrong. If your facts are clear and consistent across the web, they're easy to repeat. If they're vague or out of date, a competitor with clearer pages has the advantage. ## The prompts that decide SaaS shortlists These patterns usually decide whether you're considered. Each maps to a page type. | Prompt type | Example | What the assistant needs | Page that answers it | |---|---|---|---| | Best X for Y | "best CRM for a 20-person startup" | Who each tool is for, price, standout features | Use-case page, review profiles | | Alternatives to | "HubSpot alternatives for a small B2B sales team" | Why people switch, who fits each option | Alternatives page, roundups | | X vs Y | "Notion vs Confluence for internal documentation" | Side-by-side facts, a fair verdict | Comparison page | | Integration-led | "help desk software that integrates with Slack and Jira" | Confirmed integrations and what they do | Integration pages, docs | | Pricing | "does [tool] have a free plan" | Current plans, limits, per-seat price | Public pricing page | Keep these prompts mostly unbranded: a buyer who already knows your name doesn't need the shortlist. Our guide to [choosing prompts to track](/blog/choosing-prompts-to-track) explains how to build a balanced set. ## Comparison and alternatives pages "X vs Y" and "alternatives to" prompts are high intent, and comparison pages are a common source for them. Build them for buyers first. 1. **Pick the competitors you really meet** in sales calls, not every vendor in the category. 2. **Lead with a short verdict** on who each product suits best. 3. **Add a fact table:** pricing model, free plan, key integrations, support options, sourced where you can. 4. **Say where the competitor is stronger.** A page that only praises you reads as an ad. 5. **Date it and keep it current.** A stale comparison repeats wrong facts about them, and about you. Our own [comparison pages](/vs) follow this format, using facts from each vendor's site. Resist producing a page for every variation of a query: Google's guide to [optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) says that doing so primarily to manipulate rankings or AI responses violates its scaled content abuse policy. ## Pricing, docs and integration pages assistants can use SaaS sites often hide the facts assistants need behind demos, logins or images. - **Pricing.** Publish plan names, prices, inclusions and limits as text. If you sell custom contracts, say so ("custom pricing above 50 seats") rather than nothing. A silent pricing page pushes assistants toward review sites and old articles. - **Docs.** Public documentation answers late-stage questions like "does it support SSO" or "is there an API". Keep key pages public and current. - **Integrations.** A logo wall is weak evidence. A page per major integration stating what syncs, in which direction and on which plans gives an assistant something to cite. - **Access.** OpenAI states that sites opted out of OAI-SearchBot [won't be shown in ChatGPT search answers](https://developers.openai.com/docs/bots). Our guide on [how to rank in ChatGPT](/blog/how-to-rank-in-chatgpt) covers crawler access in more detail. An llms.txt file is optional. Our free [llms.txt generator](/tools/llms-txt-generator) builds one for AI agents, but Google says you don't need AI text files to appear in its Search. ## Review sites and roundups For "best X for Y" prompts, assistants often cite software review platforms, "best tools" articles and community threads. Keep your presence there accurate: complete profiles, current pricing, reviews requested from all customers, and corrections sent to authors of outdated roundups. Our guide on [getting cited through review sites](/blog/get-cited-by-ai-review-sites) covers how to do this within each platform's rules. Start with the domains actually cited for your prompts. ## Measure share of voice against named competitors A generic visibility score tells a SaaS team little. What matters is how often you're named compared with the vendors you compete against. - **Name your competitor set** from win-loss data, not a long list. - **Fix the prompt set** so runs stay comparable. - **Measure per assistant.** A strong ChatGPT position can hide an absence in Gemini or Perplexity. - **Record position.** Being named fourth after three rivals is not the same as being the first pick. - **Pair it with sources.** When share of voice moves, the cited domains usually explain why. [AI share of voice](/blog/ai-share-of-voice) explains the calculation and how to read a score. ## How to run GEO for SaaS with Citegram Citegram is a free Chrome extension with a dashboard built for this measurement. Our [page for SaaS teams](/for/saas) shows the workflow in detail. 1. **Set up your workspace.** Enter your company and the competitors you meet in deals, and invite teammates when you're ready. 2. **Add your buyer prompts:** category, "alternatives to", "X vs Y" and integration-led questions. 3. **Run them in the real assistants.** Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, using the accounts already signed in to your browser. No API key is needed, questions count toward your own plan usage, and first results arrive in about 10 minutes. 4. **Read share of voice and positions.** Reports show share of voice per brand and per assistant, and each answer records whether and where your brand appears. 5. **Find the pages to work on.** The most cited domains and, for ChatGPT, the web searches it ran show which review sites, comparison pages and queries matter. 6. **Change one thing, then rerun** the same prompts and follow the trend. Answers come from your own accounts, so memory, personalization and location can affect them. Read trends across repeated runs rather than single answers. ## FAQ ### What is GEO for SaaS? It's generative engine optimization applied to software buying: improving how often AI assistants name your product when buyers ask for recommendations, alternatives or comparisons. It combines clear product pages, accurate review profiles and measurement against named competitors. ### How do I get my SaaS recommended by ChatGPT? Make sure OpenAI's search crawler can reach your site, publish clear pricing, comparison and integration pages, and keep your profiles accurate on the review sites ChatGPT cites. Then track the same buyer prompts over time. No one can guarantee a recommendation. ### Which prompts should a B2B SaaS company track? The questions buyers ask before they know your name: "best X for Y", "alternatives to [competitor]", "X vs Y" and use-case questions with constraints like team size, budget or integrations. Keep the set stable so trends are comparable. ### Do comparison pages help AI visibility for SaaS? They can, because "X vs Y" and "alternatives to" prompts often draw on comparison content. Fair pages with clear facts serve buyers and assistants better than one-sided ones. Avoid mass-producing near-identical variants, which Google's guidance warns against. ## See where your SaaS stands Find out whether assistants put you on the shortlist for your buyers' questions. The free plan tracks 1 prompt across all major assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # How to Measure GEO ROI: KPIs for AI Visibility Source: https://www.citegram.com/blog/geo-roi/ Updated: 2026-10-10 To measure GEO ROI, track a chain of KPIs rather than one number: visibility in AI answers first (share of voice per assistant, position, cited domains), then the business signals that can follow (AI referral traffic, branded search, pipeline), set against what the work costs. Many AI answers produce no click, so GEO ROI is a case you build from leading and lagging indicators, not a figure an analytics tool hands you. That makes reporting harder than for paid search, but not impossible. This guide covers the AI visibility KPIs worth tracking, where each one comes from, how to talk about pipeline without overclaiming, and a simple monthly template you can copy. ## Key takeaways - GEO ROI combines leading indicators (visibility in answers) with lagging ones (traffic, branded demand, pipeline). Neither half is enough alone. - Share of voice per assistant, position in the answer and cited domains are the core AI visibility KPIs, measured on a fixed prompt set. - AI referral traffic in GA4 is real but undercounted, because many AI visits arrive without a referrer and many answers get no click. - Branded search lift is an observation worth reporting, not proof that GEO caused it. - Attribute pipeline only where you have evidence, such as self-reported attribution, and report everything else as directional. ## Why GEO ROI is harder to measure than SEO In SEO, the chain from ranking to click to conversion is mostly visible. In GEO, several links in that chain are hidden: - **Answers without clicks.** A buyer can read "HubSpot and Pipedrive are good fits for a 20-person startup", pick one and never visit either site. - **Lost referrers.** Links opened from apps or copied into a browser often arrive as Direct traffic. - **Little platform reporting.** ChatGPT, Claude and Perplexity don't give brands an impressions report. Google counts AI Overviews and AI Mode inside Search Console's "Web" data, per its [AI features documentation](https://developers.google.com/search/docs/appearance/ai-features). - **Variable answers.** The same prompt can produce different answers from run to run and from account to account. - **Long B2B cycles.** A mention in March may influence a deal that closes in September. So you measure the answers directly and treat downstream numbers with care. ## The GEO KPI stack, from leading to lagging | Layer | KPI | Where it comes from | What it tells you | |---|---|---|---| | Visibility | Share of voice per assistant | Prompt tracking | How often you're named vs competitors | | Visibility | Mention rate | Prompt tracking | The share of answers that include you at all | | Visibility | Average position | Prompt tracking | Whether you lead the shortlist or trail it | | Sources | Cited domains, and your own domain's citations | Prompt tracking | Which pages shape answers, and whether yours is one | | Traffic | AI referral sessions and key events | GA4 | Visits and conversions from answers that got a click | | Demand | Branded search impressions and clicks | Search Console | Whether more people look you up by name | | Revenue | Self-reported and CRM-sourced pipeline | Forms, CRM, sales notes | Deals where AI answers played a documented part | ### Share of voice per assistant This is the core KPI: your brand's mentions divided by all tracked brands' mentions across a fixed set of prompts. Report it per assistant before you blend it, because you can lead in Perplexity and be absent in ChatGPT. Our guide to [AI share of voice](/blog/ai-share-of-voice) covers the calculation and what distorts it. ### Position in the answer Being named first in a shortlist of three is not the same as being the fifth brand in a long list. Track your average position when mentioned, alongside share of voice. ### Cited domains Cited domains explain why the other numbers move. Track two things: which domains are cited most for your prompts, and how often your own pages appear among them. If share of voice drops, the cited domains usually show which review site or roundup tipped the answer. ### AI referral traffic GA4 has an AI Assistant channel for visits from assistants such as ChatGPT and Gemini, and Google's [default channel group documentation](https://support.google.com/analytics/answer/9756891) notes that it excludes AI Overviews and AI Mode. Report sessions and key events from that channel, labeled as a minimum. Our guide to [tracking ChatGPT traffic in GA4](/blog/track-ai-traffic-ga4) shows the setup, including a regex for assistants the default list may miss. ### Branded search lift If assistants recommend you more often, some buyers will search for your name before they buy. Search Console's [branded queries filter](https://developers.google.com/search/blog/2025/11/search-console-branded-filter) splits Performance data into branded and non-branded queries, for eligible top-level properties with enough query volume. Report the trend next to your share of voice, but present it as an observation. Campaigns, events, seasonality and press can all move branded search too. ### Pipeline and revenue This is where overclaiming is most tempting. Three honest sources: - **Self-reported attribution.** Add a free-text "How did you hear about us?" field to demo and signup forms, and count answers that mention ChatGPT, Perplexity, "AI" or similar. - **AI-referred conversions.** Key events from the AI Assistant channel in GA4, passed to your CRM where possible. - **Sales notes.** Ask reps to log when a prospect says an assistant suggested you. Each undercounts. Together they give you documented cases rather than a modeled guess. ## How to calculate GEO ROI without overclaiming Start with the cost side, which is the easy part: hours spent, content and PR budget, and tool subscriptions. For the return side, sort value into three tiers: 1. **Attributed.** Revenue from deals with AI-referred conversions or self-reported AI discovery. 2. **Assisted.** Deals where AI shows up somewhere in the journey but not as the source, such as a sales note. 3. **Directional.** Share of voice, position, cited domains and branded search trends. Use only the first tier, plus clearly documented cases from the second, in the classic formula: (attributed value − cost) ÷ cost. Report the third tier beside it, not inside it. A modest, defensible figure next to a rising share of voice is more credible than a large number built on assumptions. ## A simple monthly GEO reporting template Copy this table into your report. Keep the prompt set and competitor list fixed for the whole quarter, and note any change. | KPI | Source | This month | Last month | Comment | |---|---|---|---|---| | Share of voice, per assistant | Prompt tracking | | | One row per assistant | | Mention rate | Prompt tracking | | | | | Average position | Prompt tracking | | | | | Top 5 cited domains | Prompt tracking | | | | | Answers citing your site | Prompt tracking | | | | | AI referral sessions and key events | GA4 | | | Minimum, not total | | Branded clicks | Search Console | | | Observation only | | Self-reported AI leads | CRM | | | | Add a final line listing the actions taken this month and the prompts they targeted. If you're starting from scratch, an [AI visibility audit](/blog/ai-visibility-audit) gives you the baseline. ## Common mistakes in GEO reporting - **Changing prompts mid-quarter.** New prompts change the number without anything changing in the market. See [choosing prompts to track](/blog/choosing-prompts-to-track). - **Blending assistants too early.** A single score can hide a gain in one assistant and a loss in another. - **Treating AI traffic as the whole return.** It only counts answers that got a click. - **Reporting single runs.** Answers vary, so judge trends over several runs. ## Track the visibility KPIs with Citegram Citegram covers the visibility and sources rows of the template. It does not measure traffic, conversions or Search Console data, so pair it with GA4 and Search Console for the rest. 1. **Add your fixed prompt set and competitors.** Use unbranded buyer prompts such as "best CRM for a 20-person startup". 2. **Run them in the real assistants.** The Chrome extension asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, with the accounts already signed in to your browser. No API key; questions count toward your own plan usage. 3. **Read the per-answer detail.** Each answer records whether your brand is mentioned, its position, and the sources it cited. For ChatGPT you also see the web searches it ran. 4. **Fill the template from the reports.** Share of voice per brand and per assistant and the most cited domains feed the first rows directly. 5. **Re-run every month.** Trends over repeated runs are what make the report meaningful. Answers come from your own accounts, so personalization and location can affect them. More for [agencies](/for/agencies) and [SaaS teams](/for/saas). ## FAQ ### How do you measure the ROI of GEO? Compare the cost of GEO work with the value you can attribute to it, using AI-referred conversions and self-reported attribution. Report visibility KPIs such as share of voice and cited domains alongside, as leading indicators. Don't fold directional signals into the ROI figure itself. ### What KPIs should I track for AI visibility? Start with share of voice per assistant, mention rate, average position and cited domains, measured on a fixed prompt set. Add AI referral traffic from GA4 and branded search from Search Console as downstream signals. ### How long does GEO take to show ROI? There is no fixed timeline. Assistants that search the web can pick up new pages once they're indexed, while answers based on training data change more slowly, and B2B sales cycles add months on top. If the work is landing, visibility KPIs tend to move before revenue does. ### Can GA4 measure all traffic from AI assistants? No. Visits without a referrer land in Direct, and AI Overviews and AI Mode clicks are counted in Organic Search. GA4 shows a minimum, and it can't see answers that mention you without a click. ## Start with the first row of the report Share of voice is the KPI everything else hangs on, and you can measure it today. The free plan tracks 1 prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # How to Get Cited in Google AI Overviews Source: https://www.citegram.com/blog/get-cited-in-google-ai-overviews/ Updated: 2026-10-10 To get cited in AI Overviews, a page must first be indexed and eligible to appear in Google Search with a snippet. Google says there are no additional technical requirements. Past that point, getting cited in AI Overviews comes down to ranking for the related searches Google runs behind the Overview and offering a passage specific enough to support the answer. That makes AI Overviews SEO less mysterious than it sounds, but harder to check than a ranking: an Overview appears for some queries and not others, and its sources aren't always the top organic results. This guide covers what Google says about how sources are chosen, a practical checklist, tactics to avoid and how to track citations. ## Key takeaways - A page can only be cited in an AI Overview if it is indexed and eligible to show with a snippet in Google Search. - Google says AI Overviews may run several related searches across subtopics, so a page can be cited because it ranks for one of those sub-queries. - Google says no special schema, AI text file or new markup is needed to appear in AI Overviews. - Content with first-hand experience or an expert view that goes beyond common knowledge is what Google recommends to stand out. - Overviews vary by query, location and time, so track which sources they cite on a schedule rather than checking once. ## How Google picks the sources an AI Overview cites Google doesn't publish a full selection recipe, but its documentation describes the mechanics. Everything in this section comes from Google's pages on [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) and [optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide). ### The Overview only appears when Google thinks it helps Google says AI Overviews show when its systems determine they are "additive to classic Search". Many queries never trigger one. Before working on a query, check whether it shows an Overview at all. ### Grounding in the Search index Google describes its generative features as grounded: its core ranking systems retrieve current pages from the index, and the response is built from them and links to them. A page outside the index, or one that can't show a snippet, can't be a supporting link. ### Query fan-out Google says AI Overviews and AI Mode may issue "multiple related searches across subtopics and data sources" to build a response. For "best CRM for a 20-person startup", the sub-queries might cover per-user pricing, onboarding time or a HubSpot vs Pipedrive comparison. Your page can be cited because it answers one of those, even if it doesn't rank for the head query. Our explainer on [query fan-out](/blog/query-fan-out-explained) goes deeper. ## Cited, ranked and clicked are three different outcomes | Outcome | What it means | Where you see it | |---|---|---| | Ranked | Your page appears in the classic results | Rank tracker, Search Console position | | Cited | Your page is a supporting link inside the Overview | Only by looking at the Overview itself | | Clicked | Someone visits your page from the results | Search Console clicks, analytics | Search Console counts AI Overview traffic inside the regular "Web" search type, so it tells you little about citations. If clicks are your concern, our guide to [AI Overviews and traffic](/blog/google-ai-overviews-traffic) covers measurement. This article focuses on the middle row: being one of the sources. ## A practical checklist to get cited in AI Overviews Nothing here guarantees a citation. These steps follow Google's guidance and remove the reasons a good page gets skipped. ### 1. Confirm eligibility - **Indexed.** Use URL Inspection in Search Console on the pages you want cited. - **Snippet allowed.** Check that you haven't set `nosnippet`, a restrictive `max-snippet` or `data-nosnippet` on the passages that matter. Google lists these as controls that also apply to AI features. - **Crawlable.** Google recommends allowing crawling in robots.txt and through any CDN or hosting setup. Bot protection that blocks Googlebot blocks citations too. - **Included in the setting.** Google's guide says a site must be included in Search generative AI features in Search Console. Inclusion is the default, but check nobody changed it. ### 2. Rank for the sub-queries List the questions a buyer would need answered to act on the head query: price, limits, alternatives, setup, use cases. Check whether a page of yours ranks for each. Where it doesn't, add a clear section to an existing page rather than creating a new thin one. ### 3. Write passages that can support an answer - **Answer first.** Put the direct answer in the first sentence or two under each heading. - **Be specific.** Prices, plan limits, integrations, steps, criteria. A summary can't lean on "powerful and flexible". - **Bring something non-commodity.** Google recommends content that offers "unique expert or experienced takes that go beyond common knowledge and the ordinary". Real numbers from your own product, honest trade-offs and tested comparisons qualify. Generic tips don't. - **Keep important content in text.** Pricing shown only in an image is harder to use. - **Show freshness.** Keep facts current and display an accurate updated date. ### 4. Keep facts consistent Google's best practices include keeping Merchant Center and Business Profile data current and making sure structured data matches the visible text. Your pricing page, product pages and the third-party pages describing you should state the same facts. ### 5. Look at who else gets cited Overviews often link to third-party pages: reviews, comparisons, forums. If those pages name competitors and not you, ranking your own page won't close the gap. Our guide on [getting cited through review sites](/blog/get-cited-by-ai-review-sites) covers how to earn that presence honestly. ## What not to do Google's 2026 guide is direct about several popular tactics: - **AI text files and special markup.** Google says you don't need new machine-readable files, AI text files or special schema.org markup, and that Google Search ignores llms.txt. Structured data still helps with rich results; see our article on [schema markup for AI search](/blog/schema-markup-for-ai-search). - **Chunking.** There's no requirement to break content into tiny pieces for AI. - **One page per fan-out query.** Creating separate content for query variations "primarily to manipulate rankings or generative AI responses" violates Google's scaled content abuse policy. - **Manufactured mentions.** Google warns that seeking inauthentic mentions across the web "isn't as helpful as it might seem". If you also care about AI Mode, the rules are the same, but the answers and links can differ. Our [Google AI Mode guide](/blog/google-ai-mode) explains how. ## How to track AI Overview citations with Citegram Checking Overviews by hand works for five queries, once, but not for dozens checked repeatedly. Citegram is a Chrome extension built for this part: 1. **Pick your queries.** Start with the commercial and comparison questions that matter to your pipeline. The free [AI prompt generator](/tools/ai-prompt-generator) can help you draft a list. 2. **Add them as prompts.** Citegram runs each one in Google in a background tab and records whether an AI Overview appeared and which sources it cited. 3. **Check whether you're among the sources.** Each result shows the cited sources and whether and where your brand is mentioned. 4. **Compare with other assistants.** The same prompts run in ChatGPT, Gemini, Perplexity, Claude, DeepSeek and Grok, using the accounts already signed in to your browser, with no API key. 5. **Review the most cited domains.** Reports show the domains cited most and share of voice per brand and per assistant. First results arrive in about 10 minutes. 6. **Re-run after changes.** Trends over repeated runs show whether citations move after you update a page. Treat that as an observation, not proof of cause. Because Citegram runs from your own browser, the Overviews it sees reflect your location and Google account, the same as a manual check. SEO teams can see how this fits their workflow on the [Citegram for SEO teams](/for/seo) page. ## FAQ ### How do I get my website cited in Google AI Overviews? Make sure your pages are indexed and eligible to show with a snippet, and that your site is included in Search generative AI features in Search Console. Then rank for the related sub-queries behind your target query and put clear, specific answers near the top of each section. Google says no special markup or files are required. ### Does schema markup help you appear in AI Overviews? Google says structured data isn't required for generative AI search and there's no special schema.org markup to add. Structured data still makes pages eligible for rich results. If you use it, keep it consistent with the visible content of the page. ### Why does Google cite a lower-ranking page instead of mine? AI Overviews may run several related searches, so the cited page may rank well for a sub-query you don't cover. It may also state a specific fact your page leaves vague. Compare the cited passage with your own page to see what's missing. ### How long does it take to appear in AI Overviews? There's no fixed timeline. Google has to crawl and index your updated page, and Overviews vary by query, location and time. Track the same queries on a regular schedule and judge over several weeks rather than from a single check. ## See who gets cited for your queries If you want to know whether Google AI Overviews cite your pages, and which competitors they cite instead, start with the free plan: 1 prompt across all assistants, forever. Pro adds unlimited prompts and competitors for $10 per seat per month. [See pricing](/pricing). --- # AI Crawlers and robots.txt: GPTBot, ClaudeBot, PerplexityBot Source: https://www.citegram.com/blog/ai-crawlers-robots-txt/ Updated: 2026-10-09 AI crawlers fall into three groups, and robots.txt treats them differently: training crawlers (like GPTBot and ClaudeBot) collect content for model training, search crawlers (like OAI-SearchBot, Claude-SearchBot and PerplexityBot) index pages so assistants can cite them, and user-triggered fetchers (like ChatGPT-User, Claude-User and Perplexity-User) visit a page when someone asks. Blocking GPTBot keeps your content out of OpenAI's training. It does not remove you from ChatGPT search. Blocking OAI-SearchBot does. Many sites block "AI bots" as one category, then disappear from AI answers. This reference lists each vendor's documented user agents, explains what blocking each one does, and gives copy-paste robots.txt setups for the three most common policies. All names and behaviors come from the vendors' own documentation, last checked on 9 October 2026. ## Key takeaways - Each major AI vendor separates training crawlers, search crawlers and user-triggered fetchers, and each one reads its own robots.txt rule. - Blocking a training crawler such as GPTBot or ClaudeBot doesn't, by itself, remove you from that vendor's AI search results. - Blocking a search crawler such as OAI-SearchBot or PerplexityBot can stop that assistant from showing and citing your pages. - User-triggered fetchers may not follow robots.txt: OpenAI says rules "may not apply" to ChatGPT-User, and Perplexity says Perplexity-User "generally ignores" them. - Google-Extended controls Gemini training and grounding, not Google Search. AI Overviews and AI Mode are governed by Googlebot rules. ## Training bots vs search bots vs user-triggered fetchers **Training crawlers** gather public pages that may be used to train future models. Blocking them is a training opt-out, nothing more. **Search crawlers** build the index an assistant searches when it answers with web results. If one is blocked, the assistant has less or nothing of yours to retrieve, so it can't cite your pages for prompts like "best CRM for a 20-person startup". **User-triggered fetchers** act on behalf of a person in the chat, for example when someone pastes your URL and asks for a summary. They don't crawl the web automatically. Because a human requested the page, some vendors say robots.txt may not apply. For the bigger picture on how retrieval shapes answers, see [how AI assistants choose their sources](/blog/how-ai-assistants-choose-sources). ## The AI crawler reference table Robots.txt tokens as documented by each vendor (last checked 9 October 2026). Version numbers in full user-agent strings change, so match on the token, not the full string. | Vendor | Token | Type | What the vendor says it does | |---|---|---|---| | OpenAI | `GPTBot` | Training | Crawls content that may be used to train OpenAI's foundation models | | OpenAI | `OAI-SearchBot` | Search | Surfaces websites in ChatGPT's search features | | OpenAI | `ChatGPT-User` | User-triggered | Certain user actions in ChatGPT and Custom GPTs; not used for automatic crawling | | OpenAI | `OAI-AdsBot` | Ads review | Visits only pages submitted as ads on ChatGPT, to check their safety | | Anthropic | `ClaudeBot` | Training | Collects web content that could contribute to model training | | Anthropic | `Claude-SearchBot` | Search | Indexes content to improve the quality of Claude's search results | | Anthropic | `Claude-User` | User-triggered | Fetches pages when a Claude user asks a question that needs them | | Perplexity | `PerplexityBot` | Search | Surfaces and links websites in Perplexity results; not used for training foundation models | | Perplexity | `Perplexity-User` | User-triggered | Visits a page to answer a user's question; not used for crawling or training | | Google | `Google-Extended` | Control token | Manages use of crawled content for Gemini training and grounding | Sources: [OpenAI crawlers overview](https://developers.openai.com/docs/bots), [Anthropic's crawler help article](https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler), [Perplexity crawlers](https://docs.perplexity.ai/guides/bots), and [Google's common crawlers list](https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers). Google-Extended has no user-agent string of its own: Google crawls with its usual user agents and reads the token as a permission. It covers Gemini training and grounding in Gemini Apps and Vertex AI, so it isn't purely a training switch. ## What happens when you block each one - **GPTBot.** OpenAI says disallowing it "indicates a site's content should not be used in training generative AI foundation models". OpenAI also states that each setting is independent, so you can stay in ChatGPT search while opting out of training. - **OAI-SearchBot.** Opted-out sites aren't shown in ChatGPT search answers, though they can still appear as navigational links. OpenAI notes that robots.txt changes can take about 24 hours to reach its search systems. - **ClaudeBot.** Anthropic says blocking it signals that your future materials should be excluded from training datasets. - **Claude-SearchBot.** Anthropic says blocking it prevents indexing for search, which may reduce your visibility and accuracy in Claude's search results. - **Claude-User.** Claude can no longer retrieve your pages when users ask, which may reduce visibility. - **PerplexityBot.** Perplexity recommends allowing it so your site can appear in its search results. - **Google-Extended.** Google says it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal". It won't take you out of AI Overviews or AI Mode, which use Googlebot. If you want to limit what Google shows from your pages in Search, including AI features, Google points to `nosnippet`, `data-nosnippet`, `max-snippet` and `noindex`. Those affect classic results too, so use them with care. ## Recommended robots.txt setups Under the robots.txt standard, a crawler follows the group that names it most specifically and ignores the `User-agent: *` group. If you add a group for GPTBot, repeat any general `Disallow` rules you want it to respect inside that group. ### Stay visible in AI search, opt out of training Common for B2B sites: citable pages, no training. ```text User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: Google-Extended Disallow: / User-agent: OAI-SearchBot Allow: / User-agent: Claude-SearchBot Allow: / User-agent: PerplexityBot Allow: / ``` Note the Google-Extended trade-off: it also covers grounding in Gemini Apps, so blocking it may limit how Gemini uses your content in answers. ### Full opt-out ```text User-agent: GPTBot Disallow: / User-agent: OAI-SearchBot Disallow: / User-agent: ChatGPT-User Disallow: / User-agent: ClaudeBot Disallow: / User-agent: Claude-SearchBot Disallow: / User-agent: Claude-User Disallow: / User-agent: PerplexityBot Disallow: / User-agent: Perplexity-User Disallow: / User-agent: Google-Extended Disallow: / ``` This removes you from these vendors' AI search, not just training. User-triggered fetchers may still reach pages, per OpenAI and Perplexity, and AI Overviews stay tied to Googlebot. ### Allow everything Robots.txt allows by default. If no group names these bots and your `User-agent: *` group doesn't disallow them, they can already crawl. Explicit `Allow: /` groups simply document your intent. ## How to check your current robots.txt and server logs 1. **Read the live file.** Open `yourdomain.com/robots.txt` and search for each token in the table. Check every subdomain too, since each has its own file. 2. **Check the `*` group.** A broad `Disallow` under `User-agent: *` also applies to any AI bot without its own group. 3. **Search your server or CDN logs** for the tokens (GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot and the user fetchers). Note status codes: a wall of 403s means something other than robots.txt is blocking. 4. **Verify the IPs.** User-agent strings can be spoofed. OpenAI, Anthropic and Perplexity publish IP ranges in JSON files linked from their docs. 5. **Wait, then re-check.** OpenAI and Perplexity both mention delays of up to about 24 hours. ## Beyond robots.txt: CDNs and bot protection A clean robots.txt doesn't help if your firewall blocks the bot first. CDN and WAF tools can challenge automated traffic, sometimes through one "block AI bots" setting that doesn't separate search from training. Perplexity's docs give Cloudflare and AWS WAF instructions for allowing its bots by user agent and published IP range. OpenAI's documentation likewise points to its published IP lists for allowing OAI-SearchBot. Anthropic advises against blocking its bots by IP, because the bot then can't read your robots.txt and the opt-out may not hold. Crawlability is only a precondition. Our guides to [ranking in ChatGPT](/blog/how-to-rank-in-chatgpt) and [ranking in Perplexity](/blog/how-to-rank-in-perplexity) cover what earns the citation after that. ## Confirm you're still cited with Citegram Citegram doesn't scan robots.txt or server logs. What it shows is the outcome: whether assistants still mention and cite you after a change. 1. **Pick prompts where you're cited today**, such as "best CRM for a 20-person startup", before you edit robots.txt. 2. **Run them in the real apps.** The Chrome extension asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, using the accounts already signed in to your browser. No API key, and questions count toward your own plan usage. 3. **Record the baseline.** Citegram logs whether your brand is mentioned, its position, and the sources and links each answer cited. 4. **Re-run after the change**, once the vendors' propagation window has passed, and compare trends over repeated runs. 5. **Check the cited domains.** If your pages drop out of citations while competitors stay, review your search crawler rules first. Answers come from your own accounts, so personalization and location can affect them. For a fuller review, see our [AI visibility audit](/blog/ai-visibility-audit) guide. ## FAQ ### Should I block GPTBot? Block GPTBot if you don't want OpenAI to use your content for training. OpenAI documents it as independent from OAI-SearchBot, so blocking GPTBot alone doesn't remove you from ChatGPT search. If you want to appear in ChatGPT search, keep OAI-SearchBot allowed. ### Does blocking AI crawlers hurt my SEO? Blocking AI vendors' crawlers doesn't affect Googlebot, and Google says Google-Extended is not a ranking signal for Google Search. What you can lose is visibility in AI assistants: blocking search crawlers like OAI-SearchBot or PerplexityBot can keep your pages out of their answers. ### Does Google-Extended affect Google Search or AI Overviews? Google says Google-Extended doesn't affect inclusion in Google Search and isn't a ranking signal. It manages use of your content for Gemini training and grounding. AI Overviews and AI Mode are part of Search, where Googlebot rules and snippet controls apply. ### Do AI bots respect robots.txt? Anthropic says its bots honor robots.txt directives, and OpenAI and Perplexity tell site owners to manage their search and training crawlers through robots.txt. User-triggered fetchers are the exception: OpenAI says robots.txt "may not apply" to ChatGPT-User, and Perplexity says Perplexity-User "generally ignores robots.txt rules". ## Confirm you're still visible after changes Changing robots.txt takes minutes. Checking that assistants still cite you is the part that matters. The free plan tracks one prompt across all major assistants, forever. Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # AI Visibility Audit: A Step-by-Step Template Source: https://www.citegram.com/blog/ai-visibility-audit/ Updated: 2026-10-09 An AI visibility audit is a structured check of whether AI assistants like ChatGPT, Gemini, Perplexity and Google AI Overviews mention your brand when buyers ask about your category, where you appear, and which sources the answers rely on. It turns scattered spot checks into a baseline you can compare against and a ranked list of fixes. Most teams start by typing a few questions into ChatGPT and getting a vague feeling. An audit replaces that feeling with a fixed prompt set, repeated runs, a scoring sheet and a gap list. This guide walks through the six steps, gives you the template columns, and explains how agencies can present the results to clients. ## Key takeaways - An AI visibility audit measures mentions, position, share of voice and cited sources across a fixed set of buyer prompts and assistants. - The prompt set is the most important input: mostly unbranded, written in buyer language and tagged by intent. - AI answers vary between runs and accounts, so run each prompt more than once and read results as samples. - Cited sources and the searches assistants run are where most actionable fixes come from. - Technical access for AI crawlers is a quick check that can explain a total absence from answers. - An audit is a baseline, not a one-off: re-run the same prompts on a schedule to see what changed. ## What an AI visibility audit answers (and what it can't) A good audit answers four questions. Are we mentioned for the questions buyers ask? Where do we appear relative to competitors? Which sources do assistants cite when they answer? What should we fix first? It can't tell you how many people asked those questions, how much traffic or revenue the answers drove, or exactly why an assistant chose one brand over another. Assistant vendors don't publish ranking factors. Treat the audit as an observation of answers, and pair it with analytics if you need traffic numbers. ## Step 1: Build the prompt set Start with 15 to 30 prompts across the buying journey. Using a B2B CRM as the example: - **Category discovery:** "What is the best CRM for a 20-person startup?" - **Use case and segment:** "CRM with good email automation for a small B2B agency" - **Comparison and alternatives:** "HubSpot vs Pipedrive for a small sales team", "Alternatives to Salesforce for startups" - **Problem-led:** "How do I stop leads falling through the cracks?" - **Branded checks:** a few prompts like "Is [your brand] good for startups?" to see how you're described Pull wording from sales calls, support tickets and Search Console, not a brainstorm. Tag each prompt with intent, segment and priority, and freeze the wording. Our guide on [choosing prompts to track](/blog/choosing-prompts-to-track) covers this step in depth. ## Step 2: Run prompts across assistants Pick the assistants your buyers use. For most B2B audits that means ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, plus DeepSeek or Grok if they matter in your market. You can do this manually or with a tool. Manually, open each assistant, paste the prompt, and copy the full answer and its sources into your sheet. It works for a small set but gets slow: 20 prompts across 5 assistants, run three times, is 300 answers. Whichever route you take, note the conditions. Answers from a logged-in account can be shaped by memory, location and settings. A logged-out session or a fresh account can give a different answer. Neither is "neutral". Record which accounts you used so the next audit uses the same setup. ## Step 3: Score visibility, position and share of voice For each prompt and assistant, log three numbers: 1. **Visibility rate.** The share of runs where your brand is mentioned. Mentioned in 2 of 3 runs is 67%. 2. **Average position.** Where you appear when the answer is a list. First of five is very different from fifth. 3. **Share of voice.** Your mentions divided by all brand mentions across the prompt set. See [how to measure AI share of voice](/blog/ai-share-of-voice) for the calculation. Then roll up by assistant and by intent. Patterns usually show up here: strong on branded prompts but absent from discovery prompts, or visible in Perplexity but not in ChatGPT. Report branded and unbranded prompts separately so branded wins don't inflate the score. ## Step 4: Map cited sources and fan-out searches into a gap list This is where most fixes come from. For every answer, list the URLs the assistant cited. Group them by domain and count how often each appears. You'll typically find three kinds of sources: - **Your own pages.** Which ones get cited, and for which prompts? - **Competitor pages.** Comparison and "alternatives" pages often get cited for comparison prompts. - **Third-party pages.** Review sites, listicles, forums and media. If a roundup cited for your top prompt doesn't include you, that's a gap. Where an assistant shows the searches it ran, add them too. ChatGPT, for example, may turn one prompt into several web searches. These [fan-out queries](/blog/query-fan-out-explained) tell you what pages you need to rank for to be in the pool of sources at all. Google says AI Overviews and AI Mode may also use query fan-out, in its documentation on [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features). ## Step 5: Check technical access If you're absent everywhere, rule out access problems before anything else. - **Google AI Overviews.** Google states that a page must be indexed and eligible to show with a snippet to appear as a supporting link, with no additional technical requirements. - **ChatGPT search.** OpenAI's [crawler documentation](https://developers.openai.com/docs/bots) says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers. Blocking GPTBot is a separate, training-related choice. - **Other assistants.** Check robots.txt rules for each vendor's crawlers, and whether a CDN or bot-protection layer blocks them. Our [guide to AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt) lists the user agents. Bing Webmaster Tools also has an [AI Performance report](https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview), in public preview, showing citations of your pages in Microsoft Copilot and Bing AI answers. It's a useful extra column for the audit. ## Step 6: Prioritize fixes and set a re-audit cadence Rank each gap by two factors: how much the prompt matters to revenue, and how much effort the fix takes. Typical fixes include getting listed on a frequently cited review site, publishing an honest comparison page, adding clear facts (who it's for, price, integrations) to product pages, or unblocking a crawler. Then set a cadence. A monthly re-run of the same prompts, with the same setup, shows whether fixes moved anything. Change one thing at a time where you can, and judge by trends across runs rather than one answer. ### Audit template Use one row per prompt, per assistant, per run: | Column | What to record | |---|---| | Prompt | Exact wording, frozen | | Intent / segment | Discovery, use case, comparison, problem-led, branded | | Assistant | ChatGPT, Gemini, Perplexity, Claude, AI Overviews… | | Run date and account | When, and which account or session | | Mentioned | Yes / no | | Position | Rank in the list, or blank | | Competitors named | In order of appearance | | Cited sources | Full URLs | | Your pages cited | URLs from your domain | | Fan-out searches | Where the assistant shows them | | Gap and action | What's missing, proposed fix, owner | ### For agencies: presenting the audit to clients Lead with three numbers: visibility rate on unbranded prompts, share of voice against named competitors, and the top cited domains. Then show the gap list, ranked, with owners and dates. Be explicit about variance and account setup so clients don't read one answer as a verdict, and promise a re-audit date rather than a ranking. ## How to run steps 2 to 4 with Citegram Steps 2 to 4 are the slow part, and they're what Citegram automates: 1. **Add your prompt set.** The free plan covers 1 prompt; Pro covers unlimited prompts. 2. **Run them in the real apps.** The Chrome extension asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, using the accounts signed in to your browser. Runs count toward your own plan usage. 3. **Read mentions, positions and sources.** For each answer it records whether your brand was mentioned, its position and the cited links. For Google it records whether an AI Overview appeared. 4. **Pull the fan-out searches.** For ChatGPT it records the web searches the assistant ran. 5. **Use the reports.** Share of voice per brand and per assistant and the most cited domains feed your gap list, with trends over repeated runs. First results arrive in about 10 minutes. Step 5 (technical access) and step 6 (prioritization) stay with you. ## FAQ ### How often should you run an AI visibility audit? Run a full audit once to set a baseline, then re-run the same prompts monthly or after major changes. Answers shift as assistants update and the web changes, so a single audit ages quickly. Keep the setup identical between runs. ### How many prompts do you need for a reliable audit? For many brands, 15 to 30 well-chosen prompts are enough to see patterns, as long as each is run more than once per assistant. Coverage of intents matters more than raw volume. Larger companies with several products or regions may need more. ### Can I do an AI visibility audit for free? Yes. You can run prompts by hand in each assistant and log results in a spreadsheet using the template above. Tools save time on repeated runs; Citegram's free plan tracks 1 prompt across 7 assistants. Our [comparison of AI visibility tools](/blog/best-ai-visibility-tools) covers other options. ### Why do AI answers change between runs? Assistants generate answers rather than retrieving a fixed list, and those that search the web may find different sources each time. Account memory, location and model updates also play a role. That's why audits use repeated runs and trends, not single answers. ## Run your first audit Start with the prompt closest to revenue and see where you stand across assistants today. Citegram's free plan tracks 1 prompt across all assistants, forever, and Pro adds unlimited prompts for $10/month. [See pricing](/pricing). --- # Answer Engine Optimization (AEO): A Practical Guide Source: https://www.citegram.com/blog/answer-engine-optimization/ Updated: 2026-10-09 Answer engine optimization (AEO) is the practice of making your content and your brand easy for answer engines, such as Google AI Overviews, ChatGPT and Perplexity, to find, trust and use when they write a direct answer to a question. In plain terms, AEO is SEO aimed at the answer instead of the blue link: the goal is to be the source an assistant quotes, or the brand it names. The term competes with GEO, LLMO and "AI SEO", and the vocabulary has moved faster than the substance. This guide sorts out the names, summarizes what Google's 2026 guide on generative AI search actually says, and gives you an AEO checklist that doesn't rely on myths. For the deeper playbook on assistants, see our guide to [generative engine optimization](/blog/what-is-generative-engine-optimization). ## Key takeaways - Answer engine optimization means improving how often answer engines quote, cite or recommend you when they answer a question. - AEO, GEO, LLMO and "AI SEO" describe largely the same work, and the label matters far less than what you measure. - Google's 2026 guide says optimizing for generative AI search is "still SEO" and lists llms.txt, chunking and special schema among the things you can ignore for Google Search. - What holds up is unglamorous: answer-first pages, original content, authentic third-party mentions and pages that crawlers can reach. - You can only judge AEO by checking the answers themselves, repeatedly, across the assistants your buyers use. ## What is answer engine optimization? An answer engine is any system that responds to a question with an answer rather than a list of links. Google's featured snippets were an early version. AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Claude are the current ones. The term was long used for featured snippets and voice assistants, and came back once assistants started writing full answers with sources. Take a buyer who asks "what's the best CRM for a 20-person startup?". There are three outcomes you can aim for: - **Mentioned.** Your brand appears in the answer, next to HubSpot or Pipedrive. - **Recommended.** The answer names you as a good fit, with a reason. - **Cited.** One of your pages, or a page about you, is listed as a source. AEO is the set of actions that make those three outcomes more likely. ## AEO vs GEO vs LLMO vs "AI SEO": one discipline, many names Vendors, agencies and analysts each picked a label. They overlap so much that arguing about them is rarely a good use of time. | Term | Stands for | Usually emphasizes | |---|---|---| | AEO | Answer engine optimization | Being the direct answer: snippets, AI Overviews, assistant answers | | GEO | Generative engine optimization | Being mentioned and cited in AI-generated answers | | LLMO | Large language model optimization | How models describe your brand, including from training data | | AI SEO | (informal) | SEO practices adapted to AI search features | The terms also compete for attention. Similarweb reports that global searches for "generative engine optimization" reached about 54.3K in January 2026, against about 30K for "answer engine optimization" ([Similarweb](https://aisearch.similarweb.com/blog/aeo-vs-geo/)). That is a measure of vocabulary, not of importance. The practical conclusion: pick the word your team and clients already use, and define it by its metrics. If you want the comparison with classic search, our article on [SEO vs GEO](/blog/seo-vs-geo) covers what changes and what doesn't. ## What Google's 2026 guide says about AEO and GEO In May 2026, Google published [a guide to optimizing for generative AI features on Google Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide). It names AEO and GEO directly and is blunt about them: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." The guide lists several things you can ignore for Google Search: - **llms.txt and other "special" files or markup.** Not needed to appear in Search. - **Chunking.** No need to split content into tiny pieces for AI. - **Rewriting for AI.** No need to write in a specific way for generative AI search. - **Inauthentic mentions.** Seeking them "isn't as helpful as it might seem." - **Special structured data.** Not required, and there's no special schema.org markup for AI. See [schema markup for AI search](/blog/schema-markup-for-ai-search). It also warns that creating separate pages for every variation of how people might search, primarily to manipulate rankings or AI responses, violates Google's scaled content abuse policy, and that you should be wary of third-party tools that promise ranking success. What it recommends instead is "non-commodity content that's helpful, reliable, and people-first", plus the technical SEO best practices you already know. One limit: this guide speaks for Google Search only. ChatGPT, Perplexity and Claude run their own systems, though many of them search the web before answering, so the same foundations apply. ## Where answers come from Each answer engine builds its answer differently, which is why AEO can't be reduced to one tactic. | Answer engine | Where the answer comes from | What you can see | |---|---|---| | Featured snippets | A passage from a page that ranks for the query | The quoted page | | Google AI Overviews and AI Mode | Google's index; Google says they may run several related searches (query fan-out) | Linked sources | | ChatGPT | Training data, plus web search when it decides to search | Cited sources when it searched | | Perplexity | A live search for most questions | Numbered sources | For Google, the [AI features documentation](https://developers.google.com/search/docs/appearance/ai-features) states the bar clearly: a page must be indexed and eligible to be shown with a snippet, and there are no additional requirements. For the other assistants, the common thread is retrieval. Our explainer on [how AI assistants choose sources](/blog/how-ai-assistants-choose-sources) walks through that pipeline. ## An AEO checklist that holds up ### Answer-first sections and clear headings Start each important section with the sentence that answers the question, then give the detail. Use headings that match how people ask. This isn't "writing for AI", which Google says you don't need to do. It's writing for a busy reader, and a page that serves that reader is also easy to quote. ### Non-commodity content If your page says what twenty others say, an answer engine has no reason to pick yours. Add what only you can: first-hand testing, real pricing details, original data, clear opinions on who your product is not for. For a CRM vendor, an honest page on "Pipedrive vs HubSpot for a 20-person team", with its own comparison criteria, beats a generic "what is a CRM" post. ### Third-party mentions and entity consistency For "best X" questions, assistants often lean on review sites, comparison articles and forum threads. Earn your place there through real customers, reviews and PR, not planted posts. Then keep your facts consistent everywhere: the same category, pricing, positioning and product names on your site, your profiles and the directories that list you. ### Technical access for AI crawlers None of the above matters if crawlers can't reach your pages. Check that robots.txt doesn't block the search crawlers you care about, that key content renders as text without heavy client-side JavaScript, and that important pages are indexed. Google considers all existing technical SEO best practices still worthwhile. ## How to measure AEO Rankings tell you little about answers. AEO needs its own metrics: 1. **Visibility.** For a fixed set of buyer prompts, how often is your brand mentioned? 2. **Position.** When you're mentioned, are you first in the shortlist or fifth? 3. **Citations.** Which pages does each assistant cite, and are any of them yours? 4. **Share of voice.** How often you're mentioned compared with competitors. Our guide to [AI share of voice](/blog/ai-share-of-voice) explains how to calculate it. For Google, Search Console counts AI Overviews and AI Mode traffic in the Performance report under the "Web" search type, and Google's guide points to the newer generative AI performance report where it is available. For other assistants there's no console. You have to ask the prompts yourself, several times, because answers vary between runs, accounts and locations. ## How to measure your AEO with Citegram Citegram is a free Chrome extension, with a dashboard, built for the answer side of the work: 1. **Add the prompts your buyers ask.** Start with category and comparison questions, not branded ones. 2. **Run them in the real apps.** Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, using the accounts already signed in to your browser. No API key is needed, and questions count toward your own plan usage. 3. **Read mentions, positions and sources.** Each answer is saved with whether your brand was mentioned, its position and the sources and links it cited. 4. **Check the searches behind the answers.** For ChatGPT, Citegram records the web searches it ran. For Google, it records whether an AI Overview appeared and which sources it cited. 5. **Track the trend.** Reports show share of voice per brand and per assistant and the most cited domains, with trends over repeated runs. Answers come from your own accounts, so personalization and location can shape them. Keep the setup stable between runs. First results arrive in about 10 minutes. ## FAQ ### Is AEO replacing SEO? No. Google's own guide describes optimizing for generative AI search as "still SEO", and many assistants search the web before answering. AEO extends SEO to a new surface, the answer, and adds new metrics such as mentions and citations. ### Is AEO the same as GEO? In practice, mostly yes. AEO grew out of featured snippets and voice search, while GEO was coined for generative AI answers, but both aim to get your brand quoted, cited or recommended in answers. Pick one term and define it by what you measure. ### How do I know if my AEO is working? Track a fixed set of buyer prompts across the assistants your audience uses and log whether you're mentioned, at what position and which sources are cited. Repeat the runs over weeks, because single answers vary. For Google, add Search Console data on AI features. ### Do I need an AEO agency? Not necessarily. Most AEO work is good SEO, content and PR, which many in-house teams already do. Be cautious with anyone who promises guaranteed placement in AI answers: Google's guide warns about tools that promise ranking success. ## Measure your answer visibility Whatever you call it, AEO only works if you check the answers. The free plan tracks one prompt across all major assistants, forever, and Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # Best AI Visibility Tools (2026): GEO Trackers Compared Source: https://www.citegram.com/blog/best-ai-visibility-tools/ Updated: 2026-10-09 AI visibility tools run the questions your buyers ask in assistants like ChatGPT, Gemini, Perplexity and Google AI Overviews, then record whether your brand is mentioned, where it appears and which sources are cited. The best AI visibility tool for you depends on three things: which engines it covers, how it collects answers, and how many prompts you can track for your budget. **Disclosure: we make Citegram, one of the tools in this list.** We kept the comparison to facts each vendor publishes on its own website and link to each pricing page so you can check. Prices and features checked on October 9, 2026. Vendors change plans often, so confirm before you buy. ## Key takeaways - AI visibility tools measure mentions, position, cited sources and share of voice across a fixed set of prompts, not traffic or rankings. - The biggest differences between tools are engine coverage, how answers are collected, prompt limits and price. - Some vendors say they capture answers from the assistants' consumer interfaces rather than APIs, because answers can differ between the two. - Entry prices range from free one-off checks to custom enterprise contracts, and some engines are paid add-ons. - A two-week pilot on your own prompts tells you more than any comparison table, including this one. ## What an AI visibility tool actually measures ### Mentions vs citations vs position A **mention** is your brand name appearing in the answer. A **citation** is a link to a source the assistant used, which may or may not be your site. **Position** is where you appear in a list of recommendations. Asked "best CRM for a 20-person startup", an assistant might name HubSpot first, Pipedrive second, and cite a review site that ranks both. Those are three different signals. ### Share of voice, sources and fan-out queries Across many prompts, mentions roll up into [AI share of voice](/blog/ai-share-of-voice): how often you appear compared with competitors. Cited sources show which pages shape the answer. Some assistants also run web searches before answering, and those searches, known as [query fan-out](/blog/query-fan-out-explained), show what you need to rank for. Google states that AI Overviews and AI Mode may use this technique in its page on [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features). ## How we compared the tools - **Engine coverage (and what costs extra).** We list the engines each vendor names on its own site, and flag add-ons. Your buyers may use assistants you aren't tracking. - **Data method: real app UI vs API sampling.** A tool can call a model's API or ask the question in the consumer app, the way a buyer would. App answers can include web search and citations that an API call may not. We describe a tool's method only where the vendor states it. - **Prompt limits, refresh frequency, price.** Limits decide how much of your market you cover; price decides whether you can run long enough to see a trend. For sizing, see [how to choose prompts to track](/blog/choosing-prompts-to-track). ## Comparison table: the tools at a glance | Tool | Engines named by the vendor | Answer collection (vendor's statement) | Entry price | |---|---|---|---| | Citegram | ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok, Google AI Overviews | Real apps, through your own signed-in accounts | Free (1 prompt), Pro $10/month | | Profound | Up to 9 on Enterprise, incl. ChatGPT, Perplexity, AI Mode, Gemini, Copilot, Claude, DeepSeek | Consumer experience, not API outputs | Free trial, Enterprise custom | | Peec AI | 3 models per plan (ChatGPT, AI Overviews, AI Mode, Copilot, Gemini…); more as add-ons | UI scraping | From $95/month | | Otterly.AI | ChatGPT, AI Overviews, Perplexity, Copilot; Claude, AI Mode, Gemini as add-ons | Not stated | From $29/month | | Semrush AI Visibility Toolkit | ChatGPT, Google AI Mode | Not stated | $99/month | | Ahrefs Brand Radar | AI Overviews, AI Mode, Perplexity, Copilot, Gemini, ChatGPT; Claude available | Not detailed | Custom prompts from $50/month | | AthenaHQ | Free tier: 5 engines; Starter: 11 incl. Claude, Grok, DeepSeek | Not stated | Free tier, Starter $295/month | | SE Visible (SE Ranking) | ChatGPT, Gemini, AI Mode, Perplexity, AI Overviews | Not stated | From $99/month | | HubSpot AI Search Grader | GPT-5.4 mini, Perplexity, Gemini | One-off analysis | Free | ## The tools, one by one ### Citegram (our product) Citegram is a free Chrome extension with a dashboard. It asks your prompts in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, in background tabs, using the accounts already signed in to your browser. It records mentions, position, cited sources, the web searches ChatGPT ran, and whether an AI Overview appeared. Free covers 1 prompt forever; Pro is $10/month for unlimited prompts. The trade-off: answers come from your own accounts, so they are real app answers, but memory, location and settings can personalize them. Runs also count toward your own plan usage. Citegram does not track Google AI Mode or Microsoft Copilot. ### Profound Profound says it "captures responses directly from the consumer experience, not API outputs". Its [pricing page](https://www.tryprofound.com/pricing) lists a free trial on ChatGPT, Gemini and Google AI Overviews with a recommended prompt set, and an Enterprise plan with custom pricing and up to 9 answer engines. It suits larger teams that want daily tracking and enterprise support. ### Peec AI Peec AI's plans let you choose 3 models among ChatGPT, Google AI Overviews, Google AI Mode, Microsoft Copilot, Gemini and Naver AI. More models, including Perplexity, are paid add-ons, and the Enterprise plan opens the full list. Its documentation says it uses UI scraping. Starter is listed at $95/month (€85 in the EU) for 50 prompts with unlimited users; see the [pricing page](https://peec.ai/pricing) and our [Citegram vs Peec AI comparison](/vs/peec-ai). ### Otterly.AI Otterly's [pricing page](https://otterly.ai/pricing) lists Lite at $29/month for 15 prompts, Standard at $189/month for 100 and Premium at $489/month for 400. Every plan tracks ChatGPT, Google AI Overviews, Perplexity and Microsoft Copilot. Claude, Google AI Mode and Gemini are add-ons priced per plan. A free trial is offered. ### Semrush AI Visibility Toolkit Semrush lists the [AI Visibility Toolkit](https://www.semrush.com/kb/1493-ai-toolkit) at $99/month, with 25 prompts for Prompt Tracking on the base plan, and names ChatGPT and Google AI Mode as tracked platforms. It fits teams that already work in Semrush. ### Ahrefs Brand Radar [Ahrefs Brand Radar](https://ahrefs.com/brand-radar) offers custom prompts from $50/month, refreshed daily, and an AI Visibility Index from $199/month built on prompts modeled from Ahrefs' keyword database. Platforms shown include AI Overviews, AI Mode, Perplexity, Copilot, Gemini and ChatGPT, with Claude available. It also covers YouTube and Reddit. ### AthenaHQ AthenaHQ's [pricing page](https://www.athenahq.ai/pricing) lists a free Essential tier covering ChatGPT, Perplexity, Google AI Overviews, Gemini and Copilot, and Starter at $295/month, which adds engines including Google AI Mode, Claude, Grok, DeepSeek, Meta AI and Mistral. Usage is counted in credits, one credit per AI response. ### SE Ranking (SE Visible) SE Ranking's standalone [SE Visible](https://visible.seranking.com/) lists Basic at $99/month for 200 prompts, Core at $189/month and Plus at $355/month, tracking ChatGPT, Gemini, AI Mode, Perplexity and AI Overviews, with a 10-day free trial. SE Ranking also sells AI tracking as an add-on to its SEO platform. ### HubSpot AI Search Grader HubSpot's [AI Search Grader](https://www.hubspot.com/aeo-grader) is free and gives a one-off score out of 100 across sentiment, presence, brand recognition, share of voice and market competition. It currently analyzes GPT-5.4 mini, Perplexity and Gemini. It's a snapshot rather than ongoing tracking of your own prompts. ## Which tool fits which team - **Solo founder.** Start free: HubSpot's grader for a snapshot, Citegram's free plan to track your most important prompt over time. - **SaaS marketing team.** You need 20 to 50 prompts, several competitors and trends. Compare tools on the engines your buyers use and the cost per prompt. - **SEO lead.** If you already pay for Semrush or Ahrefs, test their AI modules first so AI data sits next to keyword data. - **Agency or enterprise brand.** Workspaces, exports and support matter more. Profound, AthenaHQ, Peec AI and Ahrefs list agency or enterprise options. For first-party citation data, Bing Webmaster Tools added an [AI Performance report](https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview) in public preview, showing how often your pages are cited in Microsoft Copilot and Bing's AI answers. ## How to run a 2-week pilot on your own prompts 1. **Pick 10 to 20 prompts** your buyers actually ask, mostly unbranded. 2. **Shortlist two or three tools** that cover the assistants your audience uses. 3. **Run the same prompts in each tool** for two weeks, without changing wording. 4. **Spot-check by hand.** Ask a few prompts yourself and compare with what each tool recorded. 5. **Compare outputs you'll act on**: mentions, competitors, cited domains, fan-out searches. 6. **Decide on cost per useful prompt**, not headline features. For a fuller method, use our [AI visibility audit template](/blog/ai-visibility-audit), then [track your brand across AI assistants](/blog/track-brand-across-ai-assistants) over time. ## How to pilot with Citegram 1. **Install the Chrome extension** and sign in to the assistants you want to track. 2. **Add your highest-priority prompt** on the free plan, or your full list on Pro. 3. **Run it.** Citegram asks the prompt in all 7 assistants in background tabs. First results arrive in about 10 minutes. 4. **Read mentions, positions and sources** for each assistant, plus ChatGPT's web searches. 5. **Repeat during the pilot** and compare share of voice and most cited domains across runs. ## FAQ ### Are AI visibility tools accurate? They record the answers they collect, but AI answers vary between runs, accounts and locations. Treat any single result as a sample and judge trends across repeated runs. Tools that state how they collect answers are easier to interpret. ### Is there a free AI visibility tool? Yes. HubSpot's AI Search Grader gives a free one-off score. Citegram's free plan tracks 1 prompt across 7 assistants with no time limit. Several paid tools, including Profound, Otterly.AI and SE Visible, offer free trials. ### What's the difference between a GEO tool and an SEO tool? An SEO tool tracks rankings and traffic for keywords on search results pages. A GEO tool tracks whether AI assistants mention, rank and cite your brand for full prompts. Some SEO suites now include GEO modules, so the line is blurring. ### How many prompts should I track? For many brands, 10 to 30 well-chosen prompts are enough to see patterns. Cover discovery, use-case, comparison and problem-led questions, and add more once you know what you'll do with the data. ## Try it on your own prompts The quickest way to judge any AI visibility tool is to run your own buyer question through it. Citegram's free plan tracks 1 prompt across all assistants, forever, and Pro adds unlimited prompts for $10/month. [See pricing](/pricing). --- # Google AI Mode SEO: How to Show Up in AI Mode Answers Source: https://www.citegram.com/blog/google-ai-mode/ Updated: 2026-10-09 Google AI Mode SEO is the work of getting your pages used as supporting links in AI Mode, Google's conversational search experience. According to Google, there are no special requirements: a page must be indexed and eligible to show with a snippet, and the usual SEO best practices apply. What changes is how AI Mode finds pages, through many related searches run in parallel, and how you measure the result. AI Mode handles longer, more complex questions than a classic results page, such as "best CRM for a 20-person startup that sells to enterprises, with a built-in dialer". This guide covers how AI Mode works, what Google says you need, and which controls and reports you have. ## Key takeaways - AI Mode is a conversational mode of Google Search for questions that need exploration, reasoning or comparisons, with follow-up questions and links to supporting pages. - To be eligible as a supporting link, a page must be indexed and eligible to appear with a snippet. Google says no special files, schema or technical requirements are needed. - AI Mode and AI Overviews both may use "query fan-out", but they may use different models and techniques, so their answers and links vary. - Creating pages for every fan-out variation to manipulate AI responses violates Google's scaled content abuse policy. - Search Console counts AI Mode in the "Web" search type, has dedicated Generative AI performance reports and offers a setting to include or exclude your site. ## What is Google AI Mode? Google [introduced AI Mode](https://blog.google/products/search/ai-mode-search/) in March 2025 as an opt-in experiment in Search Labs, describing it as a mode that "expands what AI Overviews can do". It writes an answer with links to learn more and takes follow-up questions. Google has since rolled it out widely across countries and languages. Google's documentation on [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) describes AI Mode as "particularly helpful for queries where further exploration, reasoning, or complex comparisons are needed". That is the territory of B2B buying questions: shortlists, comparisons and "which tool for my situation" prompts. ## AI Mode vs AI Overviews | | AI Overviews | AI Mode | |---|---|---| | Where it appears | Above classic results on a normal search | A separate conversational mode of Search | | When it appears | Only when Google judges it adds value, so often not at all | When the user opens AI Mode | | Conversation | A single summary on the results page | Follow-up questions in the same conversation | | Typical questions | Getting the gist of a topic quickly | Exploration, reasoning, complex comparisons | | Query fan-out | May be used, per Google | May be used, per Google | | Models and links | May differ from AI Mode | May differ from AI Overviews | Being cited in an AI Overview doesn't mean being cited in AI Mode for the same question. Google states that the two "may use different models and techniques, so the set of responses and links they show will vary." For AI Overviews specifically, see our article on [AI Overviews and traffic](/blog/google-ai-overviews-traffic). ## How AI Mode chooses links: index, snippets and query fan-out Google's guide to [optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) says these features are "rooted in our core Search ranking and quality systems". It describes two techniques. ### Grounding in the Search index Google retrieves relevant pages from its index with its core ranking systems, then generates a response from them and links to them (retrieval-augmented generation, or grounding). A page that isn't indexed can't be a supporting link. ### Query fan-out Google defines query fan-out as "a set of concurrent, related queries generated by the model to request more information". In its example, "how to fix a lawn that's full of weeds" might fan out to "best herbicides for lawns" and "remove weeds without chemicals". For a CRM buyer, a question about the best CRM for a 20-person startup might fan out into searches about per-user pricing, integrations or comparisons with Pipedrive and HubSpot. Your page can be cited because it ranks for one of those sub-queries. Our [query fan-out explainer](/blog/query-fan-out-explained) goes deeper. ## What Google says you need, and don't need, to appear **What you need:** - **Indexing and snippet eligibility.** "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet." - **Inclusion in the Search generative AI setting.** Google's guide adds that a site must be included in Search generative AI features in Search Console. Inclusion is the default. - **SEO fundamentals.** Crawling allowed in robots.txt and by your CDN, internal links, a good page experience, important content in text form and structured data that matches the visible page. **What you don't need:** - New machine-readable files, AI text files such as llms.txt, or special schema.org markup. - "Chunking" your content into tiny pieces, or rewriting pages specifically for AI. Google also warns that creating separate content for every fan-out query, "primarily to manipulate rankings or generative AI responses", violates its scaled content abuse spam policy. That rules out the "one page per sub-question" tactic. Google's conclusion is that this work is "still SEO", which matches what we describe in [SEO vs GEO](/blog/seo-vs-geo). ## How to improve your chances in AI Mode Nothing guarantees a link, but these steps follow Google's guidance: 1. **Check the basics.** Confirm your key pages are indexed and show a snippet in classic results, using URL Inspection in Search Console. 2. **Write non-commodity content.** Google contrasts generic "tips" content with pages that bring first-hand experience or a unique point of view. For B2B, that means real pricing, real limits and honest comparisons. 3. **Cover subtopics on the pages you already have.** Answer pricing, integration and onboarding questions on your pricing, integrations and product pages rather than creating thin pages for each variation. 4. **Keep facts consistent.** Pricing and features should match across your site, your Merchant Center or Business Profile data where relevant, and the third-party pages that describe you. 5. **Look after page experience.** Pages that display well on all devices, load quickly and make the main content easy to find. For a multi-assistant framework, see our guide to [answer engine optimization](/blog/answer-engine-optimization). ## Controls and reporting ### Snippet controls and the site-level setting Google lists `nosnippet`, `data-nosnippet`, `max-snippet` and `noindex` to limit what's shown from your pages in Search, including AI features. These also affect classic results. In 2026 Google added a Search Console setting called "Search generative AI". According to Search Console Help, it covers AI Overviews, AI Mode and generative AI features in Discover, it's set per property, the default is Include, and as of August 31, 2026 it's available to all websites worldwide. Google says it isn't used as a ranking signal elsewhere in Search, and it doesn't affect AI training, which Google-Extended handles. ### What Search Console and Analytics show - **Performance report.** Sites appearing in AI Overviews and AI Mode are counted in the Performance report, within the "Web" search type. - **Generative AI performance reports.** Launched in June 2026 and available worldwide since August 31, 2026, they show impressions in AI features on Search and Discover by page, country, device and date. Google says it plans to add metrics over time. - **Google Analytics.** GA4's default channels put visits from AI Overviews and AI Mode in Organic Search, not in the AI Assistant channel. See our guide to [tracking AI traffic in GA4](/blog/track-ai-traffic-ga4). None of these shows what AI Mode said about you or which competitors it recommended. ## Track the Google AI answers you can measure with Citegram To be clear: Citegram does not track Google AI Mode. It tracks Google AI Overviews, plus Gemini, ChatGPT, Perplexity, Claude, DeepSeek and Grok. AI Mode has to be checked manually. A practical routine: 1. **Write the prompts.** List 10 to 20 complex questions buyers ask, the kind AI Mode is designed for. 2. **Check AI Mode by hand.** Ask each question a few times and log mentions, competitors and links. 3. **Track the same prompts in Citegram.** It asks each prompt in the assistants above in background tabs, using the accounts already signed in to your browser, with no API key. For Google, it records whether an AI Overview appeared and which sources it cited. First results arrive in about 10 minutes. 4. **Compare cited domains.** Reports show share of voice per brand and per assistant and the most cited domains over repeated runs. Domains that recur across assistants are worth checking in AI Mode too. 5. **Pair it with Search Console.** Use its Generative AI reports for impressions, and your log for what answers say. Answers come from your own accounts, so personalization, memory and location can affect them, in Citegram and in manual checks alike. ## FAQ ### How do I get my site in Google AI Mode? Make sure your pages are indexed and eligible to show with a snippet, and that your site isn't excluded in the Search generative AI setting in Search Console. Google says no special files, schema or optimizations are required; the usual SEO work applies. ### Can I opt out of Google AI Mode? Yes. The "Search generative AI" setting in Search Console lets you exclude a property from AI Overviews, AI Mode and generative AI features in Discover. You can also use `nosnippet`, `data-nosnippet`, `max-snippet` or `noindex`, but those also limit how your pages appear in classic results. ### Does AI Mode show up in Search Console? Yes. AI Mode data is included in the Performance report under the "Web" search type. Search Console also has dedicated Generative AI performance reports showing impressions in AI features by page, country, device and date. ### Does Google AI Mode reduce clicks? It depends on the query and the site. Google says clicks from results pages with AI Overviews tend to be "higher quality", but your own data is the best guide. Compare Search Console impressions and clicks with conversions in your analytics over time. ## See where you stand in AI answers AI Mode needs manual checks, but AI Overviews and the other major assistants can be tracked automatically. The free plan lets you track one prompt across all of them. Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # How to Rank in ChatGPT: Get Your Brand Recommended Source: https://www.citegram.com/blog/how-to-rank-in-chatgpt/ Updated: 2026-10-09 To rank in ChatGPT, make sure OpenAI's search crawler can reach your pages, publish clear answers to the questions buyers ask, get mentioned on the third-party pages ChatGPT reads, and measure whether it names you for real buyer prompts. There is no ranking position to buy or claim. What you can do is improve the odds that ChatGPT finds, trusts and repeats accurate information about your brand. "Ranking in ChatGPT" is shorthand for something more specific: being the brand ChatGPT mentions when someone asks "what's the best CRM for a 20-person startup?" This guide covers how ChatGPT builds those answers, what OpenAI actually documents, and a five-step process to improve your visibility without relying on myths. ## Key takeaways - OpenAI publishes no list of ranking factors for ChatGPT. The one documented requirement is crawler access: sites that block OAI-SearchBot are not shown in ChatGPT search answers. - When ChatGPT searches the web, it often turns one prompt into several searches. Those searches, not your keyword list, decide which pages it reads. - Third-party pages such as review sites, roundups and forums often shape recommendation answers as much as your own site. - Clear, factual pages (who it's for, what it costs, how it compares) are easier for any assistant to reuse than marketing copy. - Measure first, change one thing, then measure again. Single answers vary too much to judge from. ## Mentioned, cited or recommended: pick the outcome you want Decide which result you're after first. Each calls for different work. | Outcome | What it looks like | What usually drives it | |---|---|---| | Mentioned | Your brand name appears in the answer | Being widely discussed in your category | | Cited | Your page appears as a source link | A page that directly answers the search ChatGPT ran | | Recommended | You're on the shortlist, ideally near the top | Consistent positive descriptions across many sources | For most B2B teams, the goal is "recommended" on category and use-case prompts. A buyer who reads "HubSpot or Pipedrive are good fits" without your name has already narrowed the list. ## How ChatGPT builds an answer ChatGPT answers either from what the model learned in training or by searching the web and summarizing what it finds. Questions about products and vendors often trigger search, especially when they imply something current, like pricing. ### When ChatGPT searches the web When it searches, ChatGPT shows sources alongside the answer. According to OpenAI's [overview of its crawlers](https://developers.openai.com/docs/bots), OAI-SearchBot is the crawler used for search, and "sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers." OpenAI's [publishers and developers FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq) also says any public website can appear in ChatGPT search, and refers to URLs it can obtain from a third-party search provider. OpenAI doesn't fully document which search providers it uses or how results are ranked, so treat any claim that "ChatGPT just uses Bing" with caution. What's safe to say: a page that is crawlable, indexed widely and directly relevant to the search has a better chance of being read. ### Fan-out queries: the searches you actually need to rank for A prompt like "best CRM for a 20-person startup with good email automation" often becomes several searches. ChatGPT can rewrite it into several narrower queries, such as "best CRM for small teams 2026" or "CRM email automation comparison". This is called query fan-out, and we explain it in detail in [query fan-out explained](/blog/query-fan-out-explained). The practical point: the pages ChatGPT reads are the ones that rank for those rewritten searches. If you only track your own keywords, you can miss the queries that actually feed the answer. ## Step 1: Make sure ChatGPT can reach your pages OpenAI uses three different agents, and they do different jobs: - **OAI-SearchBot** crawls for ChatGPT search. Blocking it means your site won't be shown in ChatGPT search answers, although OpenAI says it can still appear as a navigational link. - **GPTBot** collects content that may be used to train OpenAI's models. Disallowing it signals that your content should not be used for training. It is a separate decision from search visibility. - **ChatGPT-User** visits pages when a user's action triggers it, such as asking ChatGPT to open a link. OpenAI says robots.txt rules may not apply to it and that it isn't used to decide what appears in search. Check robots.txt and your CDN or firewall rules, since bot protection can block crawlers by accident. OpenAI notes that robots.txt changes can take about 24 hours to take effect. For a full walkthrough of every AI crawler, see our guide to [AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt). ## Step 2: Be the clearest answer on your own site Assistants reuse facts they can extract and attribute. Your pages should state them plainly. - **Answer-first pages.** Open product and category pages with one or two sentences saying what the product is, who it's for and what problem it solves. - **Honest comparison and alternatives pages.** "Your brand vs competitor" and "competitor alternatives" pages match the way buyers phrase prompts. Keep them fair: say where the competitor is the better fit. One-sided pages are less useful to readers and to any system summarizing them. - **Pricing clarity.** If your pricing is hidden behind a demo request, an assistant has to rely on what other sites say about it, which may be outdated. - **Specific use cases.** A page for "CRM for real estate teams" matches prompts with that context. Google's guidance on [optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) frames this work as SEO and warns against creating pages for every query variation. The same restraint makes sense for ChatGPT. ## Step 3: Get mentioned where ChatGPT looks For "best X for Y" prompts, the sources ChatGPT cites are often not vendor sites at all. They tend to be review platforms, "best tools" roundups, community threads and press coverage. 1. **Run your buyer prompts and note the cited domains.** These are the pages already shaping the answer. 2. **Check how you appear on each one.** Are you listed? Is your description current and accurate? 3. **Close the gaps.** Update your review profiles, contact authors of roundups that are missing you or outdated, and take part in community discussions honestly. Our guide on [getting cited through review sites and roundups](/blog/get-cited-by-ai-review-sites) covers this in depth, and [how AI assistants choose the sources they cite](/blog/how-ai-assistants-choose-sources) explains why these pages carry so much weight. ## Step 4: Keep your brand facts consistent everywhere If your site says "teams of any size", a review site says "best for enterprises" and an old roundup lists a pricing tier you dropped, the assistant has to reconcile conflicting signals. The result can be vague or inaccurate. - **Align the basics:** category, target customer, pricing model, key integrations, founding facts. - **Update the profiles you control:** review platforms, marketplace listings, company directories, social bios. - **Fix outdated third-party claims** where you can, starting with the domains ChatGPT cites most for your prompts. ## Step 5: Measure, change one thing, re-measure ChatGPT answers vary between runs, phrasings and accounts, so you need a baseline. 1. **Pick 10 to 30 buyer prompts.** Category, use-case and comparison questions, not your brand name. 2. **Run each one several times** and record mentions, position, competitors and cited sources. Our guide on [how to check if ChatGPT recommends your brand](/blog/how-to-check-if-chatgpt-recommends-your-brand) details the method. 3. **Change one thing**, such as publishing a comparison page or fixing a review profile. 4. **Re-run the same prompts** after a few weeks and compare trends. ChatGPT also adds `utm_source=chatgpt.com` to referral links, according to OpenAI's publisher FAQ, so you can follow visits in your analytics. Our guide to [tracking AI traffic in GA4](/blog/track-ai-traffic-ga4) shows how. ## Track your ChatGPT visibility with Citegram Citegram is a free Chrome extension with a dashboard that handles the measurement side: 1. **Add your buyer prompts** in the dashboard. 2. **Run them in the real ChatGPT app.** Citegram asks each prompt in background tabs using the account already signed in to your browser. There's no API key, and questions count toward your own plan usage. Answers come from your account, so memory and location can influence them. 3. **See whether ChatGPT named you and where.** Each run records whether your brand was mentioned and its position. 4. **Read the sources and the searches.** Citegram records the links ChatGPT cited and the web searches it ran, which shows you which pages to appear on and which queries to target. 5. **Compare with other assistants.** The same prompts run in Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews, with share of voice per brand and per assistant and trends over repeated runs. First results arrive in about 10 minutes. ## FAQ ### Can you pay to rank in ChatGPT? OpenAI doesn't document any way to pay for a place in ChatGPT's answers. Visibility there comes from being reachable by ChatGPT's search, relevant to the question and well described across the web, which is what the steps above work on. ### How long does it take for ChatGPT to mention my brand? There is no set timeline. Crawler access changes can take effect within about a day, according to OpenAI, but earning mentions on third-party pages and changing how your brand is described takes longer. Measure regularly and look at the trend over several weeks. ### Does ChatGPT use Google or Bing results? OpenAI says ChatGPT search uses its own crawler, OAI-SearchBot, and refers to third-party search providers without listing them all in detail. Strong visibility in traditional search engines helps, because ranking for the searches ChatGPT runs increases the chance your page is read. It isn't a guarantee of a mention. ### Why does ChatGPT recommend my competitor and not me? Usually because the sources it reads mention your competitor more often or more clearly. Check which domains ChatGPT cites for your prompts, see how each one describes you, and look at the searches it runs. The gap is often on third-party pages rather than on your own site. ## See where you stand today Before you change anything, find out whether ChatGPT names you for the questions your buyers ask. The free plan tracks one prompt across all assistants, forever. Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # How to Rank in Perplexity: Get Cited in Perplexity Answers Source: https://www.citegram.com/blog/how-to-rank-in-perplexity/ Updated: 2026-10-09 To rank in Perplexity, let PerplexityBot crawl your site, publish pages that answer the searches behind your buyers' questions in quotable passages, and make sure the third-party pages Perplexity cites describe you accurately. Perplexity searches the web and cites its sources next to the answer, so "ranking" there means being one of the pages it reads and links to. That visible citation list makes Perplexity one of the most transparent assistants. You can see which pages won for any prompt, including your competitors' pages. This guide covers what Perplexity officially documents about how it finds pages, a five-step process to improve your chances of being cited, and how to check where you stand. ## Key takeaways - Perplexity answers by searching the web and linking the sources it used, so being cited depends on being found for the right searches. - PerplexityBot is the crawler that surfaces and links websites in Perplexity's search results. Perplexity recommends allowing it in robots.txt and says it is not used to train AI foundation models. - Perplexity-User fetches pages when a user asks a question and, per Perplexity's docs, generally ignores robots.txt because the fetch is user-requested. - Pages that open with a direct, specific answer are easier to quote than pages that bury the facts. - Perplexity doesn't publish how it ranks sources, so measure which pages it cites for your prompts and work from that evidence. ## How Perplexity answers: live search, then cited synthesis When you ask Perplexity a question, it typically searches the web, reads a set of pages and writes an answer with numbered citations pointing to those pages. The sources are part of the product, not an afterthought, which makes Perplexity unusually easy to audit. What Perplexity doesn't publish is the detail: how many pages it considers, which index or partners it relies on beyond its own crawler, or how it weighs one source against another. Third-party guides that quote exact numbers or "ranking factors" are making educated guesses. The safest approach is to treat Perplexity like any search-based assistant: if your page isn't found for the underlying search, it can't be cited. | What Perplexity documents | What it doesn't document | |---|---| | PerplexityBot surfaces and links websites in its search results | How sources are ranked or selected | | PerplexityBot is not used to crawl content for foundation models | Which external indexes or partners it uses, if any | | Perplexity-User fetches pages for user questions and generally ignores robots.txt | How many pages it reads per answer | | Settings changes may take up to 24 hours to apply | How often a given page is recrawled | ## Step 1: Let PerplexityBot crawl you Perplexity's [crawler documentation](https://docs.perplexity.ai/guides/bots) describes two agents: - **PerplexityBot** is designed to surface and link websites in Perplexity's search results. Perplexity recommends allowing it in your robots.txt and publishes its IP ranges so you can allowlist them in your firewall. - **Perplexity-User** supports user actions: when someone asks a question, it may visit a page to help answer and link to it. Because a user requested the fetch, Perplexity says it generally ignores robots.txt rules. In practice, check three things. Your robots.txt shouldn't disallow PerplexityBot (Google's [introduction to robots.txt](https://developers.google.com/search/docs/crawling-indexing/robots/intro) is a good refresher on the syntax). Your CDN or bot-protection settings shouldn't block it either, which can happen without anyone noticing. And remember that changes may take up to 24 hours to be reflected, according to Perplexity. The Perplexity-User behavior is a policy choice worth knowing about rather than a problem to fix. If you need to restrict it, robots.txt alone won't do it. Our guide to [AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt) compares how each assistant's crawlers behave. ## Step 2: Win the searches Perplexity runs A buyer prompt like "best CRM for a 20-person startup that sells to enterprises" carries several sub-questions: which CRMs suit small teams, which handle long sales cycles, what they cost. Search-based assistants commonly break a prompt like this into several searches, a pattern known as query fan-out. Google describes the same technique for its own AI features in its [AI features documentation](https://developers.google.com/search/docs/appearance/ai-features). To work with this: 1. **List the sub-questions** behind each buyer prompt you care about. 2. **Map each one to a page** on your site that answers it directly, or note the gap. 3. **Check who ranks for those searches** in traditional search engines, since pages that are found for a query are the ones an assistant can read. 4. **Fill gaps with genuinely useful pages**, not thin variations for every phrasing. We go deeper on this in [query fan-out explained](/blog/query-fan-out-explained). ## Step 3: Make passages easy to quote Perplexity attaches citations to specific statements, so the page that gets cited is often the one with a clear sentence that supports the claim. Write with that in mind. - **Answer first.** Open each section with the direct answer, then add context. "Pipedrive's paid plans start at…" beats three paragraphs of positioning. - **Be specific.** Names, numbers, plan names, integrations and supported use cases are quotable. "Powerful and flexible" is not. - **Date what changes.** Put an "updated" date on pricing, comparison and "best of" pages, and keep them current. For questions about current tools, a page that is visibly maintained is more useful to readers. - **Use structure.** Clear headings, short paragraphs, comparison tables and FAQ sections make it easy to find the passage that answers a narrow question. None of this is a trick. It's the same clarity that helps a busy reader, and it's consistent with what we describe in [how AI assistants choose the sources they cite](/blog/how-ai-assistants-choose-sources). ## Step 4: Be present on the third-party pages it cites For recommendation prompts, Perplexity often cites pages you don't own: review platforms, "best tools" roundups, community threads, industry publications. If those pages leave you out or describe you inaccurately, the answer will too. 1. **Collect the cited domains** for your buyer prompts over several runs. 2. **Rank them by frequency.** A handful of domains often appear again and again. 3. **Audit your presence on each.** Are you listed? Is the description current? 4. **Act where you can:** update review profiles, contact roundup authors with accurate information, contribute honestly to community discussions. Our guide to [getting cited through review sites and roundups](/blog/get-cited-by-ai-review-sites) covers outreach in detail. ## Step 5: Check which pages get cited, and which competitor pages do Because Perplexity shows its sources, you can audit your visibility at the page level, not just the brand level. - **Your pages cited:** which of your URLs appear, for which prompts. - **Competitor pages cited:** which of their pages win, and what those pages do that yours don't. - **Third-party pages cited:** where you need to be present. - **Brand mentions without citations:** cases where Perplexity names you but cites someone else's page about you. Run each prompt several times. Answers vary between runs, and one result is an anecdote. Then change one thing and re-run the same prompts a few weeks later to see the trend. The same discipline applies in other assistants; see [how to rank in ChatGPT](/blog/how-to-rank-in-chatgpt) for the OpenAI side. ## See Perplexity's sources for your prompts with Citegram Citegram is a free Chrome extension with a dashboard that automates the checking in step 5: 1. **Add the buyer prompts** you want to track. 2. **Run them in the real Perplexity app.** Citegram asks each prompt in a background tab using the Perplexity account already signed in to your browser. There's no API key, and questions count toward your own plan usage. Answers come from your account, so personalization and location can influence them. 3. **See whether you're mentioned and where.** Each run records whether your brand appears and its position. 4. **Review the cited sources.** Citegram records the sources and links each answer cited, so you can see which pages you need to beat or appear on. 5. **Compare across assistants.** The same prompts run in ChatGPT (where Citegram also records the web searches it ran), Gemini, Claude, DeepSeek, Grok and Google AI Overviews, with share of voice per brand and per assistant, the most cited domains and trends over repeated runs. First results arrive in about 10 minutes. ## FAQ ### How does Perplexity choose its sources? Perplexity searches the web for your question and cites the pages it used, but it doesn't publish how it ranks or selects them. In practice, a page has to be crawlable, found for the underlying searches and contain a clear passage that supports the answer. Checking which pages it cites for your own prompts is the most reliable guide. ### Does Perplexity use Google or Bing? Perplexity documents its own crawler, PerplexityBot, which surfaces and links websites in its search results. It doesn't publicly detail which other indexes or partners it may use, so treat claims that it simply relies on Google or Bing with caution. Being well indexed and visible in search generally helps. ### How do I get my website on Perplexity? Make sure PerplexityBot isn't blocked by robots.txt or by your firewall or CDN, since Perplexity recommends allowing it for your site to appear in search results. Then publish pages that directly answer the questions your buyers ask. There's no submission form that guarantees inclusion. ### How often does Perplexity update its sources? Perplexity searches the web when answering, so answers can reflect recently published or updated pages. It doesn't publish how often individual pages are recrawled. It does say changes to crawler settings may take up to 24 hours to be reflected. ## See where you stand in Perplexity Perplexity shows you exactly which pages it trusts for your prompts, so start by looking. The free plan tracks one prompt across all assistants, forever. Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # llms.txt: What It Is, Does It Work, and Should You Add One? Source: https://www.citegram.com/blog/llms-txt/ Updated: 2026-10-09 llms.txt is a proposed Markdown file, placed at the root of a website (`/llms.txt`), that gives large language models and AI agents a short, curated summary of the site with links to its most useful pages. It is a community proposal published at [llmstxt.org](https://llmstxt.org/), not an official web standard, and Google says its Search and AI features don't use it. That gap between the hype and the documentation confuses many teams. This guide covers what the file is, what Google and the main AI vendors actually say, where it can still help, how to write one, and what matters more for your AI visibility. ## Key takeaways - llms.txt is a proposal from Jeremy Howard of Answer.AI, first published in September 2024, for a Markdown file that summarizes a website for AI systems. - Google states that Google Search, including AI Overviews and AI Mode, ignores llms.txt, so adding one neither helps nor hurts you there. - The crawler documentation of OpenAI, Anthropic and Perplexity does not describe llms.txt as an input for choosing which sources to cite. - llms.txt does not control crawling or AI training. That is the job of robots.txt. - The file is cheap to create and can help AI agents and developer tools, so adding one is reasonable if you don't expect it to change AI answers on its own. ## What is llms.txt? Jeremy Howard, co-founder of Answer.AI, published the llms.txt proposal on September 3, 2024. The problem it addresses is practical: web pages are full of navigation, scripts and layout code, while a language model can only read a limited amount of text at once. An AI agent trying to understand your site benefits from a compact starting point. According to the spec, the file contains, in this order: 1. **An H1 with the site or project name.** This is the only required element. 2. **A blockquote** with a short summary. 3. **Optional free text** with more detail, using any Markdown except headings. 4. **H2 sections listing links**, written as `[name](url)` with an optional note. 5. **An "Optional" section** for secondary links an agent can skip. The proposal also suggests offering clean Markdown versions of important pages at the same URL with `.md` appended. ### What about llms-full.txt? llms-full.txt is not part of the llmstxt.org proposal. It's a companion file some sites publish with their full content in one document, mostly for developer documentation. Anthropic's and Perplexity's documentation sites both publish one. ## What Google says about llms.txt Google's guide to [optimizing your website for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) is unusually direct. In its list of things you can ignore, llms.txt comes first. You don't need AI text files "to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them." The guide adds that creating one for other services is "completely fine" and "will neither harm nor help your site's visibility or rankings in Google Search." It lists llms.txt alongside other myths such as "chunking" content and special schema markup for AI. ### The Lighthouse audit One Google tool does look at the file. In 2026, Chrome's Lighthouse added an Agentic Browsing category with an [llms.txt audit](https://developer.chrome.com/docs/lighthouse/agentic-browsing/llms-txt). Chrome's documentation calls llms.txt "an emerging convention" and marks the audit Not Applicable when the file is missing, "as providing the file is optional at the moment." The audit flags server errors, and its published code checks for an H1, at least one Markdown link and a minimum length. So Lighthouse checks that your file is well-formed for agents. It isn't a Search ranking signal, and a missing file doesn't fail it. ## Do ChatGPT, Claude and Perplexity use llms.txt? None of them documents it. OpenAI's crawler documentation explains how to manage OAI-SearchBot (search) and GPTBot (training) through robots.txt. Anthropic's and Perplexity's crawler pages also focus on robots.txt. None presents llms.txt as a signal for which sources get cited. Two things often get confused: - **Publishing vs reading.** OpenAI, Anthropic and Perplexity all publish llms.txt files for their own developer docs. That shows they find the format useful for documentation, not that their assistants read your file when answering a buyer. - **Fetching vs using.** Some articles cite server logs showing an AI crawler requested `/llms.txt`. A request means the file was fetched, not that a model used it to write an answer. Until a vendor documents otherwise, treat claims that an assistant "reads llms.txt for answers" as unverified. ## Where llms.txt can still be useful - **Documentation and developer tools.** When developers point coding assistants or agents at your docs, a clean index with Markdown pages saves them from parsing heavy HTML. This is where the format is most adopted. - **AI agents acting for users.** Chrome's audit page says that without the file, "agents may spend more time crawling the site to understand its high-level structure." - **A clarity exercise.** Summarizing what you sell, to whom and at what price is the same work that improves the pages assistants do read. Skip it if nobody will maintain it. A file with old pricing or dead links is worse than none. ## llms.txt vs robots.txt vs sitemap.xml | | llms.txt | robots.txt | sitemap.xml | |---|---|---|---| | Purpose | Curated summary and key links for AI systems | Tell crawlers which URLs they may crawl | List URLs for search engines to discover | | Status | Community proposal | Internet standard (RFC 9309) | Protocol supported by major search engines | | Format | Markdown | Plain-text directives | XML | | Controls access or training | No | Yes, for crawlers that respect it | No | | Used by Google Search | No, per Google | Yes | Yes | To allow or block specific AI crawlers, see our guide to [AI crawlers and robots.txt](/blog/ai-crawlers-robots-txt). ## How to write an llms.txt file: template and example 1. **Start with an H1 and a blockquote.** Your brand name, then one sentence on what you do and for whom. 2. **Add plain facts.** Category, target customer, starting price, key integrations. Statements, not taglines. 3. **List your most useful pages under H2s.** Product, pricing, comparisons and docs, each with a one-line note. 4. **Move secondary pages to an "Optional" section.** 5. **Serve it at `/llms.txt` with a 200 status,** then run the Lighthouse Agentic Browsing audit. 6. **Generate it if you can.** Citegram's own site does this: citegram.com publishes `/llms.txt` and `/llms-full.txt`, generated at build time from its page and blog list. Here is an example for a fictional B2B CRM: ```markdown # Northwind CRM > Northwind CRM is a sales CRM for B2B startups with 5 to 50 people. Plans start at $15 per user per month. Northwind CRM includes pipeline management, email sequences and a built-in dialer. It integrates with Gmail, Outlook and Slack. ## Product - [Features](https://northwind.example/features): pipeline, email sequences, reporting - [Pricing](https://northwind.example/pricing): plans, user limits, annual discount - [Northwind vs Pipedrive](https://northwind.example/vs/pipedrive): features and pricing side by side ## Optional - [Changelog](https://northwind.example/changelog) - [About us](https://northwind.example/about) ``` ## What moves AI visibility more than llms.txt If you want to be named when someone asks an assistant "best CRM for a 20-person startup", the leverage is elsewhere: - **Crawlable, indexable pages.** Assistants that search the web must be able to find and read your pages. - **Clear facts on the pages assistants read.** We cover why in [how AI assistants choose the sources they cite](/blog/how-ai-assistants-choose-sources). - **Third-party coverage.** Comparison questions often pull from review sites, roundups and forums. - **Structured data that matches the page.** Google says no special schema is needed for AI features, but correct markup still supports rich results. See [schema markup for AI search](/blog/schema-markup-for-ai-search). - **Measuring answers.** The broader discipline is covered in [what generative engine optimization is](/blog/what-is-generative-engine-optimization) and our guide to [answer engine optimization](/blog/answer-engine-optimization). ## Check whether llms.txt changes your AI answers with Citegram If you add the file, compare answers before and after instead of assuming an effect. Citegram is a free Chrome extension built for that: 1. **Add the prompts that matter.** A handful of buyer questions where your brand should appear. 2. **Run a baseline before you publish.** Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, using the accounts already signed in to your browser. No API key; questions count toward your own plan usage. First results arrive in about 10 minutes. 3. **Record where you stand.** For each answer, Citegram records whether your brand is mentioned, its position and the sources cited. 4. **Publish the file and rerun the same prompts** over the following weeks. Answers vary from run to run, so compare trends. 5. **Compare share of voice and cited domains** per brand and per assistant over time. You can observe changes, but you can't prove the file caused them. Answers also come from your own accounts, so personalization, memory and location can affect them. ## FAQ ### Does ChatGPT read llms.txt? OpenAI has not documented ChatGPT using llms.txt to choose sources. Its crawler documentation explains how to control OAI-SearchBot and GPTBot through robots.txt and doesn't present llms.txt as a signal. OpenAI does publish an llms.txt for its own developer docs, which is a different use. ### Will llms.txt improve my Google rankings? No. Google says Google Search ignores llms.txt files and that creating one "will neither harm nor help" your visibility or rankings. That includes AI Overviews and AI Mode. ### Is llms.txt an official standard? No. It's a proposal published at llmstxt.org in September 2024 and refined with community input. It isn't an IETF or W3C standard, and Chrome's Lighthouse documentation calls it "an emerging convention." ### Can llms.txt block AI training? No. llms.txt contains no directives and controls nothing. To limit crawling for AI training, use robots.txt rules for the relevant crawlers, such as OpenAI's GPTBot or Google-Extended. ## Test what actually changes your AI answers Before and after any change, from llms.txt to a rewritten pricing page, the useful question is whether the answers moved. The free plan lets you track one prompt across all major assistants. Pro adds unlimited prompts for $10 a month. [See pricing](/pricing). --- # Schema Markup for AI Search: What Helps and What's a Myth Source: https://www.citegram.com/blog/schema-markup-for-ai-search/ Updated: 2026-10-09 Schema markup for AI search is useful, but not in the way much of the advice suggests. Google says structured data isn't required to appear in its generative AI features and that there's no special schema.org markup for them, and no AI assistant's documentation we know of says that JSON-LD earns you citations. What schema still does well is make pages eligible for rich results in Google Search and describe your organization and products in an unambiguous format. This article separates what Google's documentation says from the myths, summarizes two published studies, lists the schema types worth having, and shows how to check whether your markup changed anything in AI answers. ## Key takeaways - Google says structured data isn't required for AI Overviews or AI Mode, and that there's no special schema.org markup for generative AI search. - Google still recommends structured data as part of SEO because it makes pages eligible for rich results. - In a controlled Ahrefs study, pages that added JSON-LD saw no meaningful lift in AI citations compared with matched pages that didn't. - FAQ rich results are no longer shown in Google Search since May 2026, so FAQPage markup is not a route to extra visibility there. - Worth having: Organization with sameAs, Article, Product or SoftwareApplication and BreadcrumbList, all matching what's visible on the page. ## What Google says: structured data isn't required for AI features Google's [guide to optimizing for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide), the same one that calls AEO "still SEO" (see our [AEO guide](/blog/answer-engine-optimization)), is direct. Among the things you can ignore, next to [llms.txt](/blog/llms-txt), it lists overfocusing on structured data: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." It adds that structured data is still a good idea as part of your SEO strategy, because it helps with eligibility for rich results. Google's AI features documentation says the only requirement to appear in AI Overviews or [AI Mode](/blog/google-ai-mode) is that a page is indexed and eligible to be shown with a snippet. It also asks you to make sure your structured data matches the visible text on the page. The [introduction to structured data](https://developers.google.com/search/docs/appearance/structured-data/intro-structured-data) describes two jobs for markup: helping Google understand a page, and enabling richer search results. It recommends JSON-LD where possible. Required properties make a page eligible for a rich result, not guaranteed to get one. ### Myths vs what the sources say | Claim you'll see | What the sources say | |---|---| | "You need schema to appear in AI Overviews." | Google: not required. Pages must be indexed and snippet-eligible. | | "There's special schema for AI." | Google: there's no special schema.org markup for generative AI search. | | "FAQ schema gets you into AI answers." | FAQ rich results stopped showing in Google Search on May 7, 2026 ([Google's changelog](https://developers.google.com/search/updates)). No Google document links FAQPage markup to AI features. | | "ChatGPT and Perplexity cite you because of your JSON-LD." | We found no OpenAI or Perplexity documentation naming structured data as a factor in which sources they cite. Unproven either way. | | "Schema is useless now." | Google still recommends it for rich result eligibility and page understanding. | ## What the studies found, and why they seem to disagree **Ahrefs, May 2026 (correlation, then a controlled test).** [Ahrefs reports](https://ahrefs.com/blog/schema-ai-citations/) that AI-cited pages were almost three times more likely to have JSON-LD than non-cited pages. It then tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 matched control pages, 30 days before and after. Relative to controls, citations moved by -4.6% in Google AI Overviews (statistically significant, unexplained by the authors), +2.4% in AI Mode and +2.2% in ChatGPT, the last two indistinguishable from zero. The authors' own caveats: every page was already cited, schema types were pooled, and the window was 30 days. **Semrush, January 2026 (correlation).** [Semrush's technical SEO study](https://www.semrush.com/blog/technical-seo-impact-on-ai-search-study/) of URLs cited by ChatGPT Search and Google AI Mode found JSON-LD on about 40% of AI Mode-cited pages and about 30% of ChatGPT-cited pages, with Organization, Article and Breadcrumb most common. The authors state that it identifies correlations, not causation. The two aren't really in conflict. Well-maintained sites tend to add schema and also publish stronger content and earn more links, so schema shows up more on cited pages. When Ahrefs compared pages that added schema with similar pages that didn't, the difference mostly disappeared. Neither study tells you what schema does for a page that isn't cited yet. ## Where schema still helps - **Rich results.** For supported types, markup can make your listing more informative in Google Search. That affects how your result looks, which is separate from whether an AI feature uses your page. - **Entity clarity.** Organization markup with `sameAs` links ties your site to your official profiles in a machine-readable way. Google says it uses structured data to understand page content. - **Product and organization facts.** Pricing, offers and company details in markup restate what's on the page in a consistent format. The catch: assistants answer from text. If your pricing only exists in JSON-LD and not in the visible page, you've written it for the wrong reader. Put facts in plain text first, then mark them up. ## The schema types worth having | Type | Where | Why | |---|---|---| | Organization (with `sameAs`, `logo`, `url`) | Homepage or about page | Identifies your company and links your official profiles | | Article (with author and dates) | Blog posts and guides | Clear authorship and publication dates | | Product or SoftwareApplication (with offers) | Product and pricing pages | Price and product facts in a standard format | | BreadcrumbList | Any page below the homepage | Shows where a page sits in your site | Here is a minimal Organization example for a fictional CRM company. It goes in a `