How to track AI traffic in GA4 and Search Console

The first question every team asks once AI answers start appearing is how much traffic they are getting from them. The honest answer has two halves. Chat assistants send referral traffic you can measure precisely. Google's own AI surfaces send traffic you cannot separate at all. Knowing which half you are looking at prevents the most common reporting mistake in this area, which is presenting a small referral number as if it were the whole story. Google's documentation on AI features in Search is explicit that clicks from AI Overviews are reported as ordinary Search clicks, and no filter changes that.

Where does AI traffic actually come from?

There are two distinct sources and they behave nothing alike.

Chat assistants such as ChatGPT, Perplexity, Copilot and Claude open your page in a browser when a user clicks a citation. That produces a normal HTTP referrer, so the visit appears in your analytics with a source you can identify, exactly like traffic from a forum or a newsletter.

Google's AI surfaces, meaning AI Overviews and AI Mode, sit inside the Google results page. A click from an AI Overview link is a Google organic click. Search Console counts it in the same bucket as any other, and GA4 sees google / organic. There is no dimension, filter or property setting that separates them, because Google does not expose one.

The practical consequence: you can report AI referral traffic to the decimal place, and you can only infer AI Overview impact from patterns. Anyone selling you an exact AI Overview click number is modelling, not measuring.

Source	Appears as	Measurable separately?
ChatGPT	Referral from chatgpt.com	Yes
Perplexity	Referral from perplexity.ai	Yes
Microsoft Copilot	Referral from copilot.microsoft.com	Yes
Claude	Referral from claude.ai	Yes
Google AI Overviews	google / organic	No
Google AI Mode	google / organic	No

How do I isolate AI referral traffic in GA4?

Build one exploration with a filter on session source, and save it. GA4's traffic source dimensions already carry everything you need, so this is configuration rather than tagging.

Create a new exploration, add Session source / medium as the dimension, add sessions, engaged sessions, average engagement time and conversions as metrics, then apply a filter where session source matches the regex below:

chatgpt|openai|perplexity|copilot|claude|gemini|bard|you\.com|phind|poe\.com|mistral|deepseek|grok

Two refinements matter. First, use session source rather than first user source, unless you specifically want acquisition rather than behaviour. Second, add a comparison against all other traffic, because the interesting finding is almost never the volume. It is that these visitors behave differently: published analyses consistently report longer sessions and higher conversion rates from AI referrals than from ordinary organic, which changes how you value a small number.

Save it as a shared exploration so the whole team sees the same definition. Half the disagreements about AI traffic in a marketing team are two people using different filters.

What can Search Console tell you about AI Overviews?

Not as much as you want, and more than nothing. Search Console reports impressions and clicks for Google Search including AI features, without a breakdown. So you look for a signature rather than a number.

The pattern to watch is impressions holding or rising while clicks fall on informational queries. That is what it looks like when a summary answers the question above your result. It is distinguishable from a ranking loss, where impressions fall too, and from a seasonality dip, where both move together across query types.

The workable method: open the Performance report, compare the last 3 months with the same period a year earlier, filter queries containing how, what, why or best, and read the CTR column. If average position is stable and CTR is down materially on that subset while your brand and product queries are flat, you are looking at AI summaries absorbing demand. Ahrefs measured a 34.5% reduction in position-one CTR when an AI Overview is present across 300,000 keywords, which is the order of magnitude to expect on the affected subset rather than across your whole account.

Our AI Overview traffic impact simulator applies that published figure to your own click totals, which is the closest you can get to a number without inventing one.

Should you track AI crawler visits in your server logs?

Yes, and it is the most underused source of information in this whole area. Analytics only fires when a human loads your page in a browser. Server logs record every request, including the AI crawlers that fetch your content to answer somebody's question.

Filter your access logs for user agents such as GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and Google-Extended. OpenAI documents its crawler identities publicly on its bots page, and Google lists its own in the crawler overview.

Three things this tells you that nothing else will. Which pages the assistants fetch most, which is a direct signal of what they consider relevant in your category. Whether a crawler is being served an error or a redirect chain, which silently removes you from answers. And whether ChatGPT-User requests spike for particular URLs, which indicates real users asking questions your page answers, right now.

If your host does not give you raw logs, the AI-visibility platforms in our AI visibility category increasingly include agent-traffic monitoring that does the same job from a script.

What should you actually report?

Reporting AI referral sessions alone makes the channel look trivial, because for most sites it currently is, in volume terms. Reporting it as three linked numbers tells the truth.

• AI referral sessions and their conversion rate, from the GA4 exploration above, compared against ordinary organic. This is your measured number.
• Share of voice on a fixed prompt panel, from a tracker or from a manual monthly check. This is your visibility number, and it moves long before traffic does.
• CTR on informational queries in Search Console, tracked over time. This is your exposure number, and it tells you what the summaries are costing you.

Together they answer the only question a stakeholder actually has: are we being found, and is it worth anything. Our guide to tracking brand mentions in AI search covers the manual prompt-panel method, which costs nothing and works today, and the tracker comparison covers the paid options from $20 a month upwards.

What are the limits of all this?

Be honest about four of them, because they will come up in the meeting.

Referrers get stripped. Some assistant apps and privacy configurations do not pass a referrer, so a share of genuine AI traffic lands in direct. Your measured number is a floor, not a total.

Assistants are non-deterministic. Two identical prompts can return different sources. Any single observation is noise; only repeated sampling on a stable prompt set is a signal.

Attribution is worse here than anywhere. Someone reads an assistant's recommendation, does not click, and searches your brand two days later. That converts as brand search, and the assistant gets no credit in any model you can build.

Nobody can separate AI Overview clicks. Vendors who claim to are modelling from rank-tracker observations of which queries show an overview. That is a legitimate estimate and it is not measurement, and the difference matters when you are setting a target.

Conclusion

Set up the GA4 referral exploration this week: it takes ten minutes and it is the only part of AI traffic you can count exactly. Then watch CTR on informational queries in Search Console as your exposure signal, and add a fixed prompt panel as your visibility signal. Volume from AI referrals is still small for most sites, and it is growing from a base of nearly zero with unusually engaged visitors, which is exactly the profile of a channel worth instrumenting before it matters rather than after.

Compare AI visibility tracking tools AI visibility tracking tools

Frequently asked questions

Sources

Every figure in this article traces back to one of these. We link them so you can check the original rather than take our summary of it.

• Google Search Central: AI features and your website (gov)
• Google Analytics Help: traffic source dimensions (gov)
• OpenAI: bots and crawler user agents (gov)
• Google Search Central: overview of Google crawlers (gov)
• Ahrefs: AI Overviews reduce clicks by 34.5% (study)

Free tools for this

Everything below runs in your browser, with no signup and nothing uploaded.

Definitions: Citation (AI) · AI Overview · LLM visibility · Zero-click search · Prompt tracking

Related reading