What queries does ChatGPT cite my website on?
You can't optimise for AI answers you can't see. Here's how to surface the actual questions ChatGPT names your domain on — and what to do with the list.
ChatGPT cites your website on a finite, discoverable set of buyer questions — not on “everything” and not at random. To find that set you run a reverse AI search: instead of typing a question and seeing who gets named, you start from your domain and read the query–domain index backwards to surface every question the model has cited or mentioned you on. A query list — not a single visibility score — is what you need, because each row is something you can act on: an answer to defend, a competitor to study, or a gap to close. The fastest way to get the list is the free Domain Check, which reads the index across ChatGPT, Gemini and Grok at once and returns the real queries with their intent and the rival domains named alongside you.
Why is this question so hard to answer normally?
With Google you can open Search Console and read the exact queries that drive impressions and clicks. AI assistants give you nothing equivalent. ChatGPT does not publish an impressions report, the answers are generated fresh each time, and the same prompt can return slightly different sources from one session to the next. So “what queries does ChatGPT cite me on?” feels unanswerable — you can only ever see one answer at a time, and only for the question you happened to type.
That is exactly the gap reverse AI search closes. The only reliable way to know the full set is to ask the models a broad, representative range of real buyer questions, record which domains each answer cites, and then look up your domain in that record. One question at a time is anecdote; the index is the map.
How do you find the actual queries?
The mechanic is an inverted index. A background worker continuously runs buyer questions through ChatGPT, Gemini and Grok, extracts the domains each model cites or mentions, and stores the question → domain links. To answer “what does ChatGPT cite me on?” you query that index by domain — the reverse direction — and get back the list of questions where your domain appears. The full mechanic is in reverse domain lookup for AI citations.
The complete set of queries you turn up is your AI citation footprint — the answer-layer equivalent of your organic keyword profile. The quickest way to see yours is to run the free Domain Check on your own domain.
What does each query in the list tell you?
A raw list of questions is useful; an annotated one is a worklist. For every query where you are cited, you want four things:
- The literal question. The exact phrasing the model was asked, so you can tell discovery questions (“best tools for X”) from problem questions (“how do I fix Y”).
- Intent. Informational, commercial or transactional — so you can sort the ready-to-buy queries to the top.
- Model coverage. Which of ChatGPT, Gemini and Grok actually named you. They frequently disagree, so “cited by ChatGPT” is not the same as “cited by AI.”
- Competitors. The other domains named in the same answer — the brands the model treats as your substitutes. That is your competitor query overlap.
Cited vs mentioned — which one am I seeing?
It matters whether ChatGPT linked to your page as a source or merely named your brand in prose. A linked citation is a referral path and a stronger trust signal; a plain mention is still valuable but behaves differently. Treat them as distinct columns, not one number. We unpack the practical difference in mention vs citation in AI.
What do I do once I have the list?
- Defend what you win. The queries you are already cited on are assets. Keep the underlying pages fresh, accurate and well-structured so you don’t lose the slot.
- Close the gap. Run a competitor’s domain through the same check. Queries where they are cited and you are not are your AI keyword gap — the highest-leverage worklist you have.
- Investigate losses. Where a rival is named on a question you should win, dig into why ChatGPT recommends your competitor and not you.
Will the list change over time?
Yes — AI answers are not static. Models update, sources shift, and a query you win today can flip next month. A one-off check is a snapshot; if a domain matters to you, you watch it on a cadence so you catch the changes. That is the difference between a free spot-check and an ongoing monitored project. Either way, start from the list, not from a score.