Free · email-gated · real API calls
AI Citation Checker — Does AI Cite Your Brand, or Just Mention It?
Being named in an answer and being cited as its source are different outcomes with different value: one is awareness, the other sends a click and survives the next model update. Enter a prompt your customers would actually ask. This runs it through ChatGPT, Google AI Overviews and Gemini at once and separates the two, listing the sources each engine drew on. One run is one sample at one moment from an API — re-run it and watch the trend rather than trusting a single answer.
Each run calls paid AI APIs and a live Google scrape, so it is email-gated and daily-capped. Results are a sample from each engine’s API at one moment — not a promise of what every user sees. Nothing but your email and the report summary is kept.
How to read the result
Every engine × prompt is one of four outcomes. The headline score is the share of measured answers that cite you, weighting a full citation above a bare mention.
| Outcome | What it means |
|---|---|
| Cited | Your own domain appears in the answer’s list of sources — the strongest result. Full weight in the score. |
| Mentioned | Your brand is named in the answer text but your domain is not linked as a source. Partial weight. |
| Absent | Neither named nor linked. Zero weight — but it is one sample, not a verdict. |
| No answer | The engine returned nothing to score — e.g. Google showed no AI Overview for that query. Excluded from the score, not counted against you. |
Which engines are checked
| Engine | How it is checked |
|---|---|
| ChatGPT (web search) | OpenAI’s Responses API with the web-search tool. Returns the answer plus the URLs it cited. This is the API surface, which is not identical to chatgpt.com. |
| Google AI Overviews | The AI Overview block scraped from a live Google search. Google shows one on only a fraction of queries, and suppresses it on most YMYL topics. |
| Gemini (grounded) | Gemini with Google-Search grounding. Returns the answer and its grounding sources. Represents Google’s model, not the AI Overview surface. |
| Perplexity | Perplexity Sonar, which returns native citations. Shown when configured. |
What this tool can and cannot tell you
It tells you whether an AI engine cited your brand for a specific prompt at a specific moment — no more, no less. Every answer here comes from a vendor API, which is not the same pipeline as the consumer ChatGPT or Gemini apps, and AI answers shift with phrasing, location, personalisation and model updates. So a single result is a sample, not a verdict: “cited” is genuinely encouraging, “absent” is not a sentence, and the honest way to use this is to check the prompts that matter to you repeatedly and watch the trend. What it cannot do is promise that being cited once means being cited always, or that any tactic guarantees a citation — an engine cites sources it finds useful, and the work of becoming one is what AEO and GEO actually are.
AI citation checker — FAQ
Does this show exactly what a user sees in ChatGPT?
No, and no tool can. This calls the engines through their APIs; the consumer apps personalise, cache, A/B test features and carry a conversation history. What you get here is a clean, repeatable sample of what the model retrieves for that prompt — comparable across runs, which the app is not.
What is the difference between cited and mentioned?
Cited means your domain is in the engine's source list for that answer. Mentioned means your brand name appears in the text with no link back. They are separated here because they are worth different amounts: a citation sends traffic and survives the next model refresh, a mention is awareness that disappears when the phrasing changes.
Why does Google AI Overviews sometimes return no answer?
Because Google did not serve an overview for that query, which is a result rather than an error. Coverage varies by query form and vertical — informational and comparison phrasings draw overviews far more often than transactional ones. Absence of an overview is not absence from one, and the report says which of the two it observed.
Why is this tool email-gated when the others are not?
Every run spends money on live API calls to three vendors. The other tools on this site fetch a public file and parse it, so they cost nothing to run and ask for nothing. This one has a per-run cost, and the email is what makes an open, unlimited version possible to offer at all.
Does being cited once mean I will always be cited?
No. Answers move with phrasing, location, time and model version, and a single citation can come from one retrieval that will not repeat. Treat one run as a data point. A trend across repeated runs of the same prompt is the thing worth reporting to anyone.
What prompts should I test?
The questions a buyer asks before they know your name — category, problem, comparison, "best X for Y". Prompts containing your brand mostly confirm that the engine can read your own website, which is a much weaker finding than it looks. The prompts that matter are the ones you can lose.
How do I get cited if I am absent?
Work backwards from the sources the engine did use, which this report lists. Those domains are the venues that engine trusts on the topic — the realistic routes are earning a place on them, or covering the specific sub-question none of them answers well. Absence is rarely a technical problem once crawler access is confirmed.
Do you store my prompts or results?
The prompt and the result URL are attached to the email you enter, because that is what lets the report reach you and what makes the run traceable. Nothing else is retained, and results are not shared with the domains you check.
One prompt answered. Now the other forty-seven questions.
The free AI Visibility Audit runs 47 checks across crawler access, llms.txt, JSON-LD, metadata, content extractability and brand-as-entity signals — the reasons an engine had for citing someone else — and emails a 0–100 score with the three fixes that move it most.
Run the free AI Visibility Audit