Why ChatGPT Recommends Your Competitor (and How to Fix It)

ChatGPT recommends your competitor instead of you because the material it answers from mentions them more often, more clearly or in a better fit for the question. That material is either the pages it finds when it searches the web or what the model learned in training. Most often your competitor appears on the listicles, review sites and threads ChatGPT reads, ranks for the searches it runs and states its category and facts more plainly, or your own pages are blocked or don’t answer the prompt’s constraint. Each cause leaves evidence you can check.

This guide is a diagnosis: for each cause, the symptom, where to find proof and what fixes it, then an order to work in.

Key takeaways

• First confirm the gap is real: run the same buyer prompts several times, with variations, before drawing conclusions from one answer.
• Then check whether the answer used web search. An answer with sources points to retrieval problems; an answer without sources points to what the model learned in training.
• Look at third-party coverage early: often the pages ChatGPT cites name your competitor and not you.
• ChatGPT’s own searches show which queries you need to rank for.
• Fix in order: access first, then your own site’s clarity, then the sources and searches the answers depend on.

Is ChatGPT really skipping your brand? Confirm the gap first

One answer where your competitor shows up in ChatGPT and you don’t proves little. Answers vary between runs, accounts and phrasings, so check that you’re seeing a pattern.

• Run each prompt at least three times, in a fresh chat, ideally logged out or with memory off. OpenAI says ChatGPT may use saved memories and an approximate location when it rewrites a prompt into search queries, so your own account can tilt results.
• Vary the prompt. Ask the head question (“best accounting software for small businesses”), then add a constraint (“for a 10-person agency”, “under $30 a month”), then an alternatives question (“alternatives to [competitor]”). Absent everywhere means a broad problem; absent in one variation means a specific one, much easier to fix.
• Ask a branded question too. “What is [your brand]?” tells you whether ChatGPT knows you at all, and whether what it knows is right.

Our 6-step test to check if ChatGPT recommends your brand covers the setup and how to build the prompt set.

Did ChatGPT search the web or answer from training?

OpenAI’s help center says ChatGPT “may search the web automatically” when a question would benefit from current information, and that answers using search may include citations. So look at the answer:

• It shows sources. ChatGPT searched, read some pages and built the shortlist from them. Your competitor won the retrieval step, and the six causes below apply.
• It shows no sources. ChatGPT answered from what the model learned in training. Models have a training data cutoff (Anthropic, for example, publishes one for each Claude model on its models overview), so a recent or little-covered brand may be thin or absent in that data, while a competitor written about for years is not.

OpenAI doesn’t document exactly when ChatGPT decides to search, so test both: run the prompt as is, then with Search selected from the tools menu. If you appear only when search is forced, the gap is likely in training data, not your web presence. Absent in both? Start with the causes below.

Why ChatGPT recommends your competitor: six causes and how to check each

1. Your competitor is on the pages ChatGPT reads

For “best X” questions, the pages ChatGPT reads are often third-party: roundups, review platforms, Reddit threads, comparison posts. If the roundup that ranks first for the query names four tools and not yours, the answer tends to follow it.

How to check: open the sources of every answer that skips you and note, for each cited page, whether it names you, names the competitor, or neither. A page cited again and again that names them and not you is your gap. An AI source gap analysis turns this into a scored list.

Fix: get onto those specific pages honestly: pitch the roundup author with facts, earn real reviews on the platforms that keep coming up, take part in the threads with a disclosed affiliation.

2. They rank for the searches ChatGPT actually runs

OpenAI says that when ChatGPT uses search partners, it “typically rewrites your query into one or more targeted queries”, and may send more specific follow-up queries after reading the first results. So the search engine usually sees shorter, rewritten queries rather than your buyer’s question, and the pages ranking for them become the candidates.

How to check: collect the queries ChatGPT ran for your prompts (Citegram records them; see query fan-out explained for other ways), search each one, and note where you and your competitor rank. You may find the competitor owns a query like “[category] for agencies” that you never targeted.

Fix: make sure you have one strong page for the topic behind each recurring query. Not one page per query: Google’s guide to optimizing for generative AI warns against creating separate content for every variation, fan-out queries included, primarily to manipulate AI responses.

3. Their site says what they are more clearly

An assistant has to place you in a category before recommending you for it. If your homepage says “the operating system for modern teams” and your competitor’s says “accounting software for small agencies”, the competitor is far easier to match to the question.

How to check: read your homepage and pricing page as a stranger. Is there one sentence with the category, the audience and the main use case? Then ask ChatGPT “what is [your brand]?” and compare.

Fix: state the category in plain words on the pages that matter, with the audience and use cases you want to be recommended for. Our guide to entity SEO for AI shows how to write that description once and keep it consistent across the web.

4. Their facts are consistent and yours aren’t

If your pricing, plan names or category differ between your site, review profiles, directories and old articles, an assistant has to reconcile conflicting statements. A competitor whose facts match everywhere is the safer pick.

How to check: ask ChatGPT factual questions about you (price, main features, who it’s for) and compare the answers with the truth. Wrong answers can trace back to an outdated page you can find in the sources.

Fix: correct your own pages first, then the third-party profiles you control. Fixing wrong AI information about your brand walks through it.

5. ChatGPT can’t reach your site

OpenAI’s crawler documentation is direct: sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers” (they can still appear as navigational links). A robots.txt rule, CDN setting or firewall can block it by accident.

How to check: read your robots.txt for OAI-SearchBot and wildcard rules, then look in server or CDN logs for OAI-SearchBot requests and their status codes. Do your pages ever appear in ChatGPT’s sources?

Fix: allow OAI-SearchBot (GPTBot, the training crawler, is a separate decision) and allow it in your bot protection. OpenAI notes changes can take about 24 hours. Our guide to AI crawlers and robots.txt has ready-made rules.

6. Your pages don’t answer the constraint in the prompt

Buyer prompts carry constraints: team size, budget, industry, existing tools, country. If the question is “accounting software for a 5-person agency under $30 a month” and your competitor publishes a clear price and an agencies page while your pricing says “contact us”, the competitor fits the question and you don’t.

How to check: for each constraint in the prompts where you’re missing, is there a page on your site that answers it explicitly, with numbers? If you appear for the head question but drop out once a constraint is added, this is likely the cause.

Fix: publish the facts the constraint needs: public pricing or a starting price, plan limits, the team sizes and industries you serve, integrations, regions. Honest comparison pages help when the prompt names a competitor.

Symptom, cause, check, fix: quick reference

What you see	Likely cause	How to check	Fix
Competitor named, answer has no sources	Training data	Re-run with Search on; ask “what is [brand]?”	Build third-party coverage; win search answers meanwhile
Same third-party pages cited in answers that skip you	Missing from cited sources (1)	Log sources per answer, mark who each names	Outreach, reviews, disclosed participation
Competitor’s pages cited for rewritten queries	Search ranking (2)	Collect ChatGPT’s searches, check rankings	One strong page per recurring topic
ChatGPT describes you vaguely or in the wrong category	Positioning (3)	Ask “what is [brand]?”	State category, audience, use cases plainly
Wrong price, features or plan names	Inconsistent facts (4)	Ask factual questions, trace the sources	Correct own pages, then profiles and listings
Your domain never appears in sources	Crawler access (5)	robots.txt, CDN rules, logs for OAI-SearchBot	Allow OAI-SearchBot
You appear for the head question, not with a constraint	Constraint not answered (6)	Compare prompt variations	Publish pricing, limits, segments, integrations

How to find out why ChatGPT skips your brand in about an hour

1. Pick five prompts where you expect to be recommended and aren’t, including one with a constraint and one “alternatives to [competitor]”.
2. Run each three times in a clean ChatGPT session, then once more with Search selected.
3. Record per answer: sources shown or not, brands named in order, your position if any, cited URLs.
4. Classify the answers: sources or no sources.
5. For sourced answers, mark each cited page as naming you, the competitor, both or neither. Sort by how often each page appears.
6. Collect the searches behind the answers and check your ranking for the three most frequent.
7. Ask the branded questions: “what is [brand]?”, “how much does [brand] cost?”, “who is [brand] for?”
8. Check robots.txt and logs for OAI-SearchBot.
9. Match your findings to the table above and pick the top two rows.

How to get ChatGPT to recommend you: fix in this order

Work from the most blocking and cheapest fix to the slowest:

1. Access. If OAI-SearchBot is blocked, nothing else matters for search answers.
2. Your own facts and positioning. Category sentence, public pricing, plan limits, segments, integrations. You control these pages and every other fix builds on them.
3. Constraint pages. Pages for the segments and use cases in your prompts, written for buyers rather than for a query list.
4. The top cited third-party sources. Start with the two or three pages that appear most often in answers that skip you. Google’s guide notes that seeking inauthentic mentions “isn’t as helpful as it might seem”, so aim for real reviews and accurate coverage.
5. Training-data presence. Sustained coverage in reputable places, for future models. It’s slow and you can’t measure it directly, so track the search side meanwhile.

Change one or two things at a time and re-run the same prompts so you can tell what worked. For the wider playbook, see how to rank in ChatGPT.

Do the diagnosis with Citegram

Logging runs by hand is where most teams stop. Citegram is a free Chrome extension with a dashboard that collects the evidence for you:

• It asks your prompts in ChatGPT, Gemini, Perplexity, Claude, Grok, Mistral, Copilot, Google AI Overviews and Google AI Mode, in background tabs, with the accounts already signed in to your browser. No API key; questions count toward your own plan usage.
• Each run records whether your brand was mentioned, its position, the competitors named and the sources each answer cited, in rank order.
• For ChatGPT, it records the web searches it ran, which is the evidence for cause 2.
• The dashboard shows share of voice per brand and per assistant, the most cited domains and trends over repeated runs, so you can see whether a fix worked.

Comparing assistants also helps the diagnosis: if Perplexity, which says it searches “the internet in real-time” (Perplexity help center), names you while ChatGPT doesn’t, the gap may be specific to ChatGPT’s sources or searches rather than your site as a whole. First results arrive in about 10 minutes.

FAQ

Why does ChatGPT recommend my competitor and not me?

Usually because the pages it reads name your competitor and not you, the competitor ranks for the queries ChatGPT runs, or their site states category, price and audience more clearly. If the answer shows no sources, it comes from training data, where a better-covered competitor has the edge.

How does ChatGPT decide which brands to recommend?

When it searches, ChatGPT rewrites the question into one or more search queries, reads pages that rank for them and builds its shortlist from what those pages say. When it doesn’t search, it answers from what the model learned in training. In both cases, brands that are named clearly and often in the material it relies on are the ones that get recommended.

Why does ChatGPT give different recommendations each time?

Answers vary between runs, accounts and phrasings. OpenAI says ChatGPT may use saved memories and an approximate location when it rewrites a prompt into searches, so your own account can tilt the results. Run each prompt several times in a fresh chat before concluding anything.

How do I know if ChatGPT searched the web for its answer?

Answers that use search may include citations and a Sources button. If you see none, ChatGPT most likely answered from what the model learned. You can force a search by selecting Search from the tools menu and compare the two answers.

Can I ask OpenAI to recommend my brand?

Not in any way OpenAI documents. Its documentation describes crawler controls, not a way to request inclusion in recommendations. What you can control is crawler access, your own pages and your honest presence on the third-party pages ChatGPT reads.

How long until ChatGPT starts recommending my brand?

It depends on the cause. A crawler fix can take effect in about a day, according to OpenAI, while changes to your own pages and to cited third-party pages take longer because they have to be recrawled and ranked again; no assistant publishes a timeline. What the model learned in training only changes when models are retrained or replaced, so expect months.

Find out why you’re missing

Run your buyer prompts, see who gets named and which sources and searches decided it. The free plan tracks 1 prompt and 3 competitors on 1 assistant with 7 days of history, forever. Pro adds 20 prompts, every assistant and 90 days of history for $19 a month. See pricing.