Guide
Why does ChatGPT recommend your competitor instead of you?
Ask ChatGPT the question your own buyers ask, and it names a rival by name. Nothing is broken. Retrieval decided this. The model read more of the web than your homepage, and the competitor’s page held a sentence direct enough for the model to lift and use. Retrieval is the process an AI assistant uses to pull specific passages out of indexed pages and drop them into an answer. ChatGPT did not weigh your product against theirs. It matched a buyer’s question against whichever page held a sentence direct enough to lift whole. A competitor’s page stating a figure that yours only implies wins that match. The product underneath was never on trial.
What makes one page quotable and the other isn’t?
Running a company’s own buyer questions through ChatGPT, Claude and Perplexity on a schedule and recording whether the company gets named is how this pattern shows up as a repeatable measurement rather than a hunch. The same two pages get tested, week after week, and one keeps losing the same question.
Trust Signals, an AI-visibility analytics publisher, documented the shape of that loss across a set of 2026 comparisons: the page that wins the citation carries a directly extractable, quotable claim, something like “reduces recovery time to under an hour,” plus corroboration from somewhere other than the company’s own site. A homepage that says “trusted leader in fast recovery” has nothing in it for a model to lift and attribute. A competitor’s page that states the number does.
The corroboration half matters as much as the sentence. A Trustpilot-commissioned study of over 800,000 AI responses across ChatGPT, Gemini, Perplexity and Google AI Mode, run with research partner Seer Interactive and published in May 2026, found citation rates for otherwise comparable brands rose from about 1% with no third-party review profile to 53.5% with even a handful of reviews. A claim on a company’s own page can be completely accurate and still go unquoted if nothing outside that domain repeats it. One source is not enough.
Is this the same problem as ChatGPT never mentioning your business at all?
No. The fix is different. Why doesn’t ChatGPT recommend my business? covers total absence: a company that doesn’t appear in an AI answer for its own category, for any version of the question. This is the sharper, second case: ChatGPT already retrieves the company’s page for some buyer questions, already knows it exists, and still reaches past it for the rival’s page on this one.
Losing zero times and losing to a named rival are different diagnoses. The first means nothing on the site is being indexed or trusted yet. The second means something specific on a rival’s page beat something specific, or missing, on this one. Fixing the second case means finding the exact question and rewriting the exact sentence, not rebuilding the whole site.
What can you actually do about a specific lost match?
Start with the buyer’s exact phrase, not the company’s whole website. Take a real query, “best [category] for [use case]” or “is [competitor] worth it,” and check whether the company’s own page answers it in one extractable sentence near the top. Most homepages don’t. They open with a mission statement and bury the specific claim three scrolls down, if it exists at all.
The check we run for a business does that directly. It lists the buyer questions the company is currently losing and names the page it lost to. Each one comes with a note on where and how to answer it on the company’s own site, and the homepage copy and FAQ that follow are written for the same purpose.
Publishing the answer is only half the job. How do I get my website recommended by ChatGPT? covers the general mechanics: crawlable formats, structured markup, machine-readable answers. This piece is about the narrower case where retrieval nearly worked and one missing sentence was the difference.
What tends to get quoted, and what gets skipped?
| Signal | Usually quoted | Usually skipped |
|---|---|---|
| Claim specificity | ”Reduces recovery time to under an hour" | "Trusted by leaders everywhere” |
| Third-party echo | Reviewed or discussed off the company’s own domain | Only appears on the company’s own domain |
| Answer location | One sentence, near the top of the page | Buried after a mission statement |
| Machine-readable format | Structured FAQ, llms.txt, JSON-LD present | Prose-only page, no markup |
Common questions
Does this mean my competitor paid ChatGPT to be recommended?
No. There is no paid placement inside a ChatGPT, Claude or Perplexity answer. Every brand named in one of these answers was retrieved from public content the model could find and trust, not from bought media.
How do I find out which specific buyer questions my company is losing?
Start from the company’s website. The check reads it and works out who the company competes with, then puts real buyer-intent questions to the assistants to record whether and how the company gets mentioned. The result is the list of specific questions still being lost, who is winning each one, and where on the site to fix it.
Is this a Google SEO problem or something different?
A different mechanism with some shared inputs. Traditional search ranks a page in a list of links. This is about whether a model can lift one self-contained sentence from a page and paste it into a generated answer. A page can rank first on Google and still carry no extractable claim, in which case the model skips it for a lower-ranked competitor page that has one.
Courage Horizon sets up AI agents for UK small businesses: software that does one job, such as chasing unpaid invoices, and asks you before it sends anything. Being found when customers ask ChatGPT is part of that work. Get a free report on your own business.