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 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 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.
Anthropic’s three bots and what blocking each one does
Anthropic’s crawler help article 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.
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.
- 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 can help you write separate rules for training and search bots. - 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.
- 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.
- 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.
- 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 covers how to show up there.
These are the same principles we describe in how AI assistants choose the sources they cite. 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.
- Choose 10 to 30 buyer prompts without your brand name: category, use-case and comparison questions. Our guide to choosing prompts to track explains how, and the AI prompt generator gives you a first list.
- Run each one several times. Answers vary between runs, and Claude may adapt localized results to the location inferred from your IP address.
- Record mentions, position, competitors and cited sources. Note which of your pages appear and which third-party pages Claude relies on.
- 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:
- Add your buyer prompts in the dashboard.
- 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.
- See whether Claude named you and where. Each run records whether your brand was mentioned and its position.
- Review the cited sources. Citegram records the sources and links each answer cited, so you can see which pages Claude trusts for your prompts.
- 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.
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.
ChatGPT
