Why does ChatGPT cite my competitor and not me?

ChatGPT cites a page, not a company.

That page may be your competitor’s own guide, or a third-party review that mentions them. Those call for different comparisons. So before asking why you lost, open the exact URL the answer linked.

A citation shows which page ChatGPT linked; it does not explain the choice. OpenAI’s help page puts it this way: “ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed.” That tells you what the ranking is for, not why that page was chosen over yours. What you can inspect is the page itself: how it answers the buyer’s question and what evidence sits beside that answer.

The comparison below takes 20 minutes. It should leave you with one supported fix, or a clear record of what you still do not know.

Two terms first. A citation is a link in the answer to a source page. A crawler is the program a search company sends to read web pages; ChatGPT’s search crawler is called OAI-SearchBot.

What happens when ChatGPT searches?

OpenAI’s search help page describes several parts of the process:

• ChatGPT typically rewrites your question into one or more targeted queries, which it sends to search partners.
• It may add your general location, based on your IP address.
• If Memory is on, it may use your saved memories when it rewrites the query.
• To be eligible, a website must let OAI-SearchBot crawl it, and its host or content delivery network (the service that serves and protects the site) must allow traffic from OpenAI’s published crawler addresses.

So two people asking “the same” question can trigger different searches. A screenshot records one result, not a stable preference. Access is a precondition, not a reason: a site that blocks the crawler is not eligible, and a site that allows it is still not guaranteed a citation. For what each OpenAI bot does, see GPTBot vs OAI-SearchBot.

Five explanations to check

Treat these as hypotheses. A difference between two pages does not prove why one was cited.

Possible explanation	What to look for	Check this first
Your page may be inaccessible to the crawler	Crawl rules or firewall responses that block it	Check robots.txt, the allowed crawler addresses, and your server logs
Their page answers the exact question; yours covers the topic	Their page opens with the answer; yours opens with a slogan	Read the first 100 words of both pages
A third party supplied the answer	The link goes to a review, comparison, or news site that mentions them	Look at the domain of the cited URL
Your answer is hard to reach	It sits behind a tab, a click, or a form, or only in a PDF	Check the page source and load the page with JavaScript switched off
The result varies between runs	A later answer cites a different page	Ask the same question 3 to 5 times in fresh chats and keep every result

The third row changes which page you study. A 2025 preprint from the University of Toronto reported that the AI search engines tested favoured earned media (independent review, comparison, and news sites) over brand-owned and social content, while Google showed a more balanced mix. The researchers queried the engines through developer APIs. Treat the finding as a reason to check third-party coverage, not as the explanation for your particular result.

If the citation leads to a comparison site, read what it says about your competitor and whether it mentions you at all. Rewriting your own homepage will not change that review.

Did they pay for it?

OpenAI says advertisers cannot shape, rank, or alter ChatGPT’s responses. Its ads help page also says ads do not influence the answers and run on separate systems. Ads appear below a response, labelled as sponsored.

That is OpenAI’s own statement, not an independent audit. Still, a citation alone is no evidence of payment. Before you compare pages, check what you are looking at: a source link inside the answer, or a sponsored placement below it.

Compare the two pages, row by row

Start from the exact URL attached to the relevant sentence in the answer, not your competitor’s homepage. A company can be cited through a support article, a comparison page, or an interview on someone else’s site.

Save the question, the answer, the date, and the cited URL. Then save both pages before anyone edits them: a screenshot and the page source (right-click, “View page source”). These show what your browser received and displayed, not what ChatGPT fetched.

Choose your page by the question it answers. If the buyer asked whether a service supports several approvers, compare your approval guide with theirs; your company history page is not a fair match. If you have no page that answers the question, that is your first finding.

Now fill in eight rows:

1. Claim: what the page actually says in answer to the question.
2. Where the answer sits: in the opening paragraph, or below a sales section.
3. Who is named: a product and company name next to the answer, or “our platform”.
4. Date: a visible revision date, if there is one. Record it without assuming newer is better.
5. Evidence on the page: steps, limits, numbers, or sources a reader can check.
6. Answer in the page source: whether the answer is there or loads later. On its own, this row says nothing about crawler access.
7. Crawler access: what your server logs show. For the cited site, write “unknown” unless you can see its logs.
8. Canonical: the URL the page names as its main version. Record whether it points to this page or a broader one; the tag does not prove which URL an engine used.

A worked example (hypothetical)

Suppose two suppliers sell shift-planning software. A buyer asks how a manager can correct an approved staff rota and keep a record of the change. ChatGPT links the competitor’s change-history guide. The comparison is illustrative; no competitor audit or citation experiment was run.

Check	You	Them
Claim	”Flexible planning for modern teams”	Approved rotas keep a change history; only managers can edit them
Where the answer sits	Procedure below a sales section	Procedure in the opening paragraph
Who is named	”Our platform”	Product and supplier named next to the steps
Date	No revision date	Revision date on the guide
Evidence on the page	A feature promise	Visible steps and a stated limitation
Answer in the page source	Procedure loads after opening a tab	Procedure in the page source
Crawler access	Your logs show a firewall “are you human?” check blocking an OAI-SearchBot request	Page publicly reachable; their logs unknown
Canonical	Points to the broad product page	Points to the guide itself

Start with the firewall block: eligibility depends on access, so better copy cannot help a page the crawler is stopped from reading. Then put your actual procedure, product name, and limitation near the top of the page, make sure the answer is in the page source, and check that the canonical points to the page you want cited. Each change fixes an observed problem. None guarantees a citation.

Do not copy their claim. Your product may handle corrections differently. The goal is your own supported answer to the same question, with its limits intact.

What if the two pages look equally good?

Then pause the rewrite. Check discovery and timing instead: when your page went live, whether other pages on your site link to it, and whether your logs show OAI-SearchBot requests for it. A logged request does not tell you which version informed an answer.

Ask the question again and keep every result. Some variation is expected, and your comparison cannot reveal every page ChatGPT considered. A tie is a finding: recording it saves you from paying for copy changes with no defect to fix.

Mistakes that waste the comparison

• Comparing brands or homepages instead of the cited URL.
• Writing “they have more authority” with nothing observable behind it.
• Marking the competitor’s crawler access as fine because they were cited.
• Changing the question and the page at the same time, then calling the next answer proof.
• Treating a name in the answer as a citation. A mention is not a citation.

Try this today (20 minutes)

1. Capture the result (3 minutes). Ask your buyer’s question in a fresh ChatGPT chat with web search on. Save the exact question, the answer, the date, and the cited URL.
2. Check whose page it is (2 minutes). Open that URL. Record whether it belongs to your competitor or a third party, and find the passage that answers the question.
3. Choose your page (3 minutes). Find your page that answers the same question, or record that none exists. Save the pages you will compare.
4. Fill in the eight rows (8 minutes). Write “unknown” wherever you lack evidence, including logs you cannot get within the 20 minutes.
5. Assign one next step (4 minutes). Choose the first supported fix and name who will make it. If there is no clear defect, assign a discovery or access check instead.

Make the change after the comparison, not during it. Then ask the same question again in fresh chats and keep the earlier results. A later citation is an observation; it does not prove your change caused it.

One comparison covers one question. To track many, my book The Findability Engine uses a step called Identify: keep a plain sheet of the queries buyers type, the engines that answer them, and the competitor each answer credits today. The book is still in production and not for sale; you can keep the sheet without it.

If your page never shows up for any question, start with the absence diagnosis and the citation-readiness audit instead.

Cite this:Why does ChatGPT cite my competitor and not me?Len P. van der Hof. https://lenvanderhof.com/en/blog/why-chatgpt-cites-your-competitor/ · Published 19 October 2026.