Does ChatGPT cite your site? How to check

There is exactly one reliable way to find out whether ChatGPT cites your website: ask it the questions your buyers actually ask, with web search turned on, and read the list of sources it returns. If your domain is not in that list, ChatGPT is not citing you for that question today. Everything else, including your Google rankings and your analytics dashboard, is a proxy that can be wrong.

Why can you not see this in your analytics?

Because a citation is not a visit. When someone asks an AI engine a question and reads the answer, your page was consulted but never opened. It reaches your logs only when the reader clicks the source link, and even then the referrer is often stripped.

That leaves most site owners with three bad signals:

• Google Search Console, which reports the classic blue links and says nothing about who is quoted inside an AI answer.
• Server logs, which show that an AI crawler fetched the page but not whether anything was cited from it.
• Brand-name searches, which almost always return your site and therefore prove nothing.

The check has to be run from the answer side, not the traffic side.

How do you check by hand?

Ten minutes and a notebook are enough for a first pass.

1. Write down five questions a potential customer would type, in their words. Four of them should not contain your brand name.
2. Open a fresh session with web search enabled, so the engine fetches sources rather than answering from memory.
3. Ask each question and read the sources panel, not the prose. The prose may name your brand while the citation goes to a competitor.
4. Record, for each question, whether your domain appears at all and in what position.
5. Repeat a week later. Answers vary between runs, so a single test is an anecdote.

Two details decide whether the result means anything. Use a signed-out session, because memory and personalisation will happily quote a site you visit daily. And check the domain, not the brand: being described in the prose while a rival is the linked source is a loss.

What questions should you ask?

The questions that matter are the ones where you are not the obvious answer. A useful set covers four shapes:

• Informational, with no brand: "what is X", "how does X work". This is where citations are won or lost.
• Navigational, with your brand: "is X any good", "X pricing". Not being cited here points at a technical problem.
• Comparison: "best tools for X", "X vs Y". Assembled from several sources, so the easiest place to appear.
• Long tail: a specific task, phrased the way a person would say it out loud.

myGEOscore builds that set automatically. It reads your title, H1, meta description and opening text, pulls the topic and strongest keywords out of them, and generates five queries per page: two informational, one navigational using the brand in your domain, one comparison and one long tail.

How does myGEOscore run the same check?

The three visibility factors in the paid report each ask a live engine the generated queries and then look for your domain in the sources that come back.

• The ChatGPT check calls OpenAI's Responses API with the web search tool enabled, and reads the url_citation annotations attached to the answer.
• The Perplexity check queries Perplexity's sonar model and reads the citation list it returns.
• The Google AI Overviews check queries Gemini with grounding enabled and reads the grounding sources behind the answer.

A source counts as yours when the citation URL matches your full URL, your domain or your hostname, so example.com/pricing and www.example.com both register. The position of the first match is kept too, because the fifth source is not the first. For the AI Overviews check, appearing in one of the five answers scores 4 out of 10, two scores 6, three scores 8, and four or more scores 10.

These three checks are the AI Platform Visibility category. They are reported on their own, beside the score rather than inside it: no page-type rubric gives them any weight, so they move the grade in neither direction. The score measures whether your page is citable, the checks measure whether it was cited. They run in the paid report only, and each one is written up on the methodology page.

What does a citation actually depend on?

The visibility checks measure the outcome. The 48 factors that do feed the score measure the causes, and those are the part you control: how extractable your content is, what structured data you expose, what authority signals you carry, how dense the page is in citable facts, whether AI crawlers can reach and render it, and how clearly your brand reads as an entity.

There is no one weighting: the page is classified first, then scored with that page type's weights. On an article, which is what this post is, content structure carries 25% and authority 25%, then citability 20%, schema 15%, technical readiness 10% and brand 5%. The methodology page prints all five columns. That ordering is not arbitrary: a page an engine cannot parse into a clean answer will not be cited, however authoritative the domain.

What do you do when the answer is no?

Work the causes in weight order rather than chasing the citation directly.

1. Make the first paragraph answer the question. Engines extract from the opening, and a page that starts with "in this guide we will explore" has given them nothing to quote.
2. Turn your headings into the questions people ask, so the engine can match a query to a section.
3. Add the facts. Numbers, dates, named sources and definitions are what gets quoted; adjectives are not.
4. Check that GPTBot, ClaudeBot, PerplexityBot and Google-Extended are not blocked in robots.txt or by your CDN. This one is binary, and it silently undoes everything else.
5. Re-check in a month. Citations move as engines re-crawl, so the only meaningful measurement is a repeated one.

Every one of those is a scored factor with its own guide in the factor library, including what we check and how the score is calculated.

The fastest way to see where you stand is to run the free check on the page you most want cited. It scores 36 factors, needs no signup, and tells you within a minute whether your citability problem is content, structure or a blocked crawler. The paid report at $19 adds the three live citation checks plus 3 monthly re-scans, which is what turns a one-off answer into a trend. Details are on the pricing page.