How to Tell If ChatGPT and Other AI Assistants Are Citing Your Site

There is no report anywhere that tells you how often ChatGPT cites your site. Google Search Console does not break AI features out, ChatGPT publishes no publisher dashboard, and the vendors selling AI-visibility tracking are running the same prompt panel you could run yourself, at scale.

What exists is five partial measurements. A prompt panel run by hand tells you what answers people actually see. Bing's AI Performance report is the only first-party citation report from any engine. Referral traffic tells you who clicked. Server logs tell you who fetched. And your domain's reference profile tells you whether you are in the pool the engines draw from at all. Run all five and you have a picture. Run one and you have a number that will mislead you.

Can you see AI citations in Google Search Console?

Not separately. Google's documentation on AI features is explicit: "Sites appearing in AI features (such as AI Overviews and AI Mode) are included in the overall search traffic in Search Console. In particular, they're reported on in the Performance report, within the 'Web' search type."

So the data is in there, mixed with everything else, with no filter to pull it out. There is one indirect read worth doing. Sort your Performance report by impressions, compare against clicks over the same period, and look for queries where impressions hold or rise while clicks fall. An AI Overview answering the query above your result produces exactly that shape. It is inference, not measurement, and the AI Overviews guide goes through reading it properly.

What does Bing Webmaster Tools show?

More, and it is the only place any engine reports citations to you directly. Microsoft launched an AI Performance report in Bing Webmaster Tools on 10 February 2026. Per the Bing Webmaster Blog, it "shows when your site is cited in AI-generated answers across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations", including which URLs get referenced and how that changes over time.

Two caveats. It covers Microsoft's surfaces only, so it says nothing about ChatGPT, Perplexity or Gemini. And it reports citations rather than clicks, which is the right thing to report but means the numbers are not comparable to a Search Console clicks figure. Verify the property and turn it on regardless. It is free and it is the closest thing to ground truth anyone publishes.

How do you find AI referral traffic in analytics?

Look for the assistant hostnames in your acquisition report. In Google Analytics 4 that is Reports, then Acquisition, then Traffic acquisition, grouped by session source. The rows to look for are chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and claude.ai. Links in ChatGPT answers also carry a utm_source=chatgpt.com parameter, which shows up in server logs and in any analytics tool that records the query string.

Three things distort this measurement, and all three push the number down:

• Some assistant traffic arrives without a referrer and lands in Direct. Your AI referrals are therefore a floor, not a total.
• A citation is not a click. Most people read the answer. A page can be cited constantly and send almost no traffic, and that is the normal outcome, not a failure.
• Bot and preview fetches inflate raw pageviews if you have not filtered them, which cuts the other way.

Treat referral traffic as evidence that citations exist, not as a count of them.

What do your server logs tell you?

Whether the engines are fetching you at all, which is the prerequisite everything else depends on. Each vendor documents its user agents, and the search crawlers are the ones that decide whether you can be cited:

An absence here is the most actionable finding in this whole guide. If OAI-SearchBot has never fetched you, no page-level work will get you cited, and the fix is in robots.txt or your CDN. Whether to block AI crawlers covers what each agent controls.

How do you run a prompt panel?

This is the measurement that reflects what people actually see, and it is manual. It is also the one the tracking vendors automate and resell, so doing it yourself for twenty questions costs an hour a month and tells you most of what a subscription would.

1. Write twenty questions your site should be the answer to, in the words a customer would use. Not keywords. Questions.
2. Run each one through ChatGPT search, Perplexity, Google AI Mode and Copilot. Use a logged-out or fresh session so personalisation and memory do not colour the result.
3. Record every cited domain, not just whether you appeared. The list of who does get cited is the real output: it is the set of sources the engines currently treat as credible in your subject.
4. Note the shape of the answer as well as the sources. Which sub-question did the engine actually answer? That tells you what to write next.
5. Repeat monthly, same questions, same wording. Answers are volatile between model releases, so a single run is noise and the trend is the signal.

Using the Revised directory for this

Step 3 produces a list of domains the engines cite in your subject. That list is the input to the measurement most people skip: how your domain's provenance compares with theirs.

Run the Agent Citability checker on your own domain and on each domain from the panel. Agent Citability is our published measurement of how deeply a domain is referenced by the kinds of sources answer engines cite. It ships as versioned releases and a score can fall between them, so read it as a snapshot. Then run the backlink checker on the same list to see who references them and with what anchors. Between the two you can usually see, concretely, why the engines pick those domains and not yours.

If the gap is large and mostly about references rather than content, the directory is the shortcut worth considering. Filter under Quality on a minimum Agent Citability score, use the source filter or a hub like expired domains with .edu backlinks to start from a reference type you cannot otherwise obtain, and read the Linked by column to check the subject matches yours. Every filter is free to use; reveals are what a plan buys. The AI visibility guide for a new website covers the whole argument, and referring domains versus backlinks covers reading the counts.

What none of this can tell you

How many times you were cited. Nobody except the vendors knows, and only Microsoft publishes any of it.

Whether a citation changed anything. There is no impression-to-outcome path for an answer somebody read and did not click.

Why an engine picked one source over another. All of the above is observation of outputs. The selection mechanism is not published by any vendor, and a measurement study of Google AI Overviews found nearly 30 percent of cited domains did not appear in the first-page results for the same query, so it is not simply ranking either.

FAQ

How do I know if ChatGPT is citing my website? Ask it. Run twenty questions your site should answer through ChatGPT search and record which domains it cites. For a rough traffic read, look for chatgpt.com in your analytics acquisition report and the utm_source=chatgpt.com parameter in your logs. There is no first-party ChatGPT report for publishers.

Does Google Search Console show AI Overview traffic? Not as its own row. Google says AI feature traffic is included in overall search traffic and reported in the Performance report under the Web search type, with no separate breakdown.

Is there a free AI visibility tracker? The free method is a prompt panel you run by hand, plus Bing's AI Performance report, which is free with a verified Bing Webmaster Tools property and is the only first-party citation data any engine publishes.

Why do I get AI citations but no traffic? Because most people read the answer. A citation is an appearance in the source list, and the click-through on that is low by design. Judge citation work on whether you are named, not on referral volume.

My logs show GPTBot but never OAI-SearchBot. What does that mean? That OpenAI has crawled you for training but its search crawler has not, so you are unlikely to appear in ChatGPT's answers. Check robots.txt and your CDN's bot rules for a rule catching OAI-SearchBot.

Once you know which domains the engines cite in your subject, the useful question is what they have that you do not. It is usually references, and it is usually years of them. The Revised directory lists expired domains that already carry them, spam-screened, with the referencing sources named on every listing before you reveal it.