How to Track AI Referral Traffic (ChatGPT, Perplexity, Gemini) in GA4

To track AI referral traffic, check GA4's Traffic acquisition report for the "AI Assistant" default channel first — Google Analytics already catches many visits from ChatGPT, Gemini and Copilot automatically. Then build one custom channel group to catch the assistants or app traffic that slip through as "Referral" or "Direct," since Google doesn't publish its full matching list and in-app browsers often strip the referrer entirely.

What counts as AI referral traffic?

AI referral traffic is a real visitor who clicked a link an AI assistant gave them in an answer — ChatGPT citing your page, Perplexity linking a source, Gemini or Copilot naming your site. It's a session in your analytics tool, the same as a visitor arriving from a search result or a tweet.

That's different from an AI crawler (GPTBot, PerplexityBot, and similar) visiting your site to index it for training or retrieval. Crawler visits show up in your server logs, not as sessions in Google Analytics, and tracking them is a separate problem from this one.

Does GA4 already track AI referral traffic?

Yes, partly. Google Analytics 4's default channel group includes an AI Assistant channel. A session is assigned to it when "the medium exactly matches 'ai-assistant'" (something you'd set manually with UTM tagging) or, more relevant here, when "the referrer matches a list of AI Assistants" Google maintains internally, according to Google's own GA4 channel definitions (checked October 2026).

Google doesn't publish that referrer list, and reports differ on which assistants are covered, so the only reliable way to know what you're getting is to check your own property.

How do you check what's already landing in the AI Assistant channel?

1. Open Reports > Acquisition > Traffic acquisition in GA4.
2. Set the primary dimension to Session primary channel group.
3. Look for a row named AI Assistant. That's traffic GA4 already classified for you.
4. Add Session source as a secondary dimension on the Referral and Direct rows too, and scan for sources like chatgpt.com, perplexity.ai, claude.ai, gemini.google.com or copilot.microsoft.com. If you see them there instead of under AI Assistant, that assistant isn't on Google's internal list for your traffic, and you'll need a custom channel to catch it reliably.

How do you build a custom channel group to catch what's missing?

A custom channel group lets you add a channel with your own rule, and GA4 applies it retroactively to your reports, per Google's instructions for custom channel groups (checked October 2026):

1. In Admin, under Data display, open Channel groups.
2. Click Create new channel group (this starts as a copy of the default group, so the native AI Assistant channel stays intact).
3. Click Add new channel, name it something like "AI referrals (manual)."
4. Add a condition: dimension Source, operator matches regex, pattern covering the domains you found in step 4 above, for example: chatgpt\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com
5. Click Reorder and drag your new channel above Referral in the list — GA4 assigns a session to the first channel whose rule matches, and Referral's own rule ("medium is referral, app or link") would otherwise catch this traffic first.
6. Save group, then set it as your property's primary channel group if you want it to be what you see by default.

Standard GA4 properties can have up to two custom channel groups beyond the default one, so this won't compete with other custom groups you've already built.

Check this list against your own Session source data every few months — assistants change domains and app behavior, and this is exactly the kind of list Google's hidden internal match can miss.

Why does some AI traffic still show up as Direct?

Because the referrer header never arrives. Many AI assistants' mobile apps and in-app browsers don't pass a referrer when a user taps a link, the same limitation that makes a lot of social-app traffic show as Direct too. A custom channel group can't fix this — there's no source value to match against. If your Direct traffic has grown noticeably since AI assistants started citing you, some of that growth is very likely AI-driven even though GA4 can't prove it channel by channel.

Some trackers report that ChatGPT appends a utm_source parameter to certain citation links, which would show up as campaign data even without a clean referrer. OpenAI hasn't published when or how consistently it does this, so don't build your whole measurement around it — check your own Traffic acquisition report for sessions with utm_source=chatgpt.com as a bonus signal, not the main one.

Does Google Search Console show AI referral traffic too?

No — it shows something related but different. Search Console's Generative AI performance report, which Google announced in June 2026 and rolled out to more sites afterward, tracks how often your pages are shown inside Google's own AI Overviews and AI Mode. Per Google's help page for the report (checked October 2026), it reports impressions only — "how many times links to your site were shown to a user in a generative AI feature" — with no click or visit data, and it covers only Google's own AI features, not ChatGPT, Perplexity, Claude or Copilot.

So the two tools answer different questions: Search Console's report tells you whether Google's AI Overviews are showing your pages at all; GA4's AI Assistant channel (plus your custom one) tells you whether anyone actually clicked through from any assistant and landed on your site.

What do you do once you can see the traffic?

Once AI referral sessions are visible as their own channel, treat them like any other source: check which pages they land on, whether those visitors convert, and whether the content on those pages is what's getting cited (clear, factual, dated and sourced pages tend to be the ones AI answers quote back). It's the same instinct behind checking which startup directories still send real traffic instead of assuming a listing helps just because it exists, or picking channels in a first-100-users plan based on what actually converts rather than what feels productive.

Larvabot's own reporting works the same way for the channels it watches directly — it attributes visits cookielessly back to the specific reply, email or directory listing that sent them, which matters once you're trying to tell "an AI assistant cited my backlinks page" apart from "someone found it through a Reddit reply." For more on this cluster, see measuring what works.

Frequently asked questions

What is referral traffic?

Referral traffic is any visit to your site where the visitor arrived by clicking a link on another site or app, as opposed to search results (organic), typing your URL directly, or a paid ad. GA4 groups referral-type visits into channels like Referral, Organic Search, Organic Social and, now, AI Assistant, based on the source and medium values it detects.

How is AI referral traffic different from AI crawler traffic?

Referral traffic is a human clicking through from an AI assistant's answer — it's a session in your analytics tool. Crawler traffic is an AI company's bot (GPTBot, PerplexityBot and similar) fetching your pages to index or retrieve from, which shows up in server logs, not in GA4, and happens whether or not any human ever visits.

Why is my ChatGPT or Perplexity traffic showing as Direct instead of Referral?

Most often because the app or in-app browser the visitor used didn't send a referrer header at all, which GA4 reads as a Direct visit with no source information to classify. This is a known limit of browser- and app-level privacy behavior, not something your GA4 setup can fix.

Do I need GA4 360 to track AI referral traffic?

No. Everything in this guide — the native AI Assistant channel and a custom channel group — is available on a standard, free GA4 property. GA4 360 only raises the number of custom channel groups you're allowed (five instead of two) and other data limits that most small sites won't hit.

Will this work if I use a tool other than GA4?

The specific steps are GA4's, but the idea applies anywhere: check whether your analytics tool already separates out AI-assistant referrers, then build a custom rule matching the referrer domains in the table above for whichever ones it's missing. Any tool that lets you segment sessions by referrer domain can do this.