To track AI traffic in GA4, create a custom channel group that groups sources like chatgpt.com, perplexity.ai and gemini.google.com into a single “AI Assistants” channel, then order it above the default Referral channel so those sessions get intercepted before they’re miscategorised. That gives you a clean, reportable view of how much traffic — and how many conversions — generative engines are actually sending you.
The catch: a large share of AI traffic never shows up as a referral at all. It lands in Direct, because most AI assistants pass no referrer when a user clicks through. So the full method is two-part — build the channel group, then interrogate your Direct traffic. This guide walks through both.
Why AI traffic hides in Direct (and why it matters)
Here’s the problem in one sentence: you almost certainly have more AI-driven traffic than your analytics is telling you.
When someone clicks a link inside ChatGPT, Perplexity or a Copilot answer, the browser often sends no referrer header. Desktop apps and in-app browsers are the worst offenders. With no referrer, GA4 has nothing to work with and files the session under Direct — the same bucket as someone typing your URL by hand.
The scale of this is not trivial. Analysis across large web datasets in early 2026 found that between 35% and 70% of AI referral sessions arrive with no referrer information and end up in Direct. If your Direct traffic has crept up over the last 18 months with no obvious cause, a chunk of it is generative search.
And the trend line only points one way. AI referral traffic sat at roughly 1.08% of all website visits in early 2026 but is projected to reach 20–28% of total referral traffic by the end of the year, according to industry tracking. ChatGPT alone drives about 64.5% of generative AI web traffic, on the back of more than 800 million weekly active users.
For anyone reporting on marketing performance, this is the gap between a defensible number and a guess. If you can’t separate AI traffic from Direct, you can’t prove that your content SEO work is earning citations, and you can’t argue for more investment in it. Measurement is what turns “AI search is important” into a budget line.
How to track AI traffic in GA4: the custom channel group
GA4 lets you define your own channel grouping rules. This is the core of the setup and it takes about ten minutes.
Step 1 — Create a new channel group
In GA4, go to Admin → Data display → Channel groups. Do not edit the Default Channel Group — GA4 uses it in too many places and you don’t want to break existing reports. Click Create new channel group and give it a clear name like “AI-aware channels”.
Step 2 — Define the AI Assistants channel
Add a new channel at the top and set the condition to match your AI sources. Use a regex on the Source dimension so one rule covers everything:
chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|claude\.ai|copilot\.microsoft\.com|bing\.com/chat|meta\.ai|bard\.google\.comName the channel “AI Assistants”. This catches every session where the source matches one of those hosts.
Step 3 — Order it above Referral
This is the step most people miss, and it’s the one that makes the whole thing work. GA4 evaluates channel rules top to bottom and assigns each session to the first channel it matches. If Referral sits above your AI Assistants channel, AI sources get swallowed by Referral first and your channel stays empty.
Click Reorder and drag “AI Assistants” so it sits just above Referral. Now AI sessions are intercepted and labelled correctly before the generic Referral rule ever sees them.
Step 4 — Add landing page as a secondary dimension
Once data flows in, build an exploration or report using your new channel group and add Landing page as a secondary dimension. This is the single most useful output of the whole exercise: it tells you which specific pages generative engines are citing and sending traffic to. Those are your proven AI-visible assets — the pages worth doubling down on.
Channel groups in GA4 apply retroactively to historical data, so you’ll see past AI traffic reclassified immediately, not just from today forward. Google documents the full behaviour in the GA4 channel groups reference.
GA4’s native AI Assistant channel — and its blind spot
In May 2026, Google added a native AI Assistant channel to GA4’s default grouping. It automatically classifies traffic from recognised assistants — ChatGPT, Gemini, Claude and others — without any setup on your part.
This is genuinely useful and worth turning to first. But do not treat it as complete. It only catches sessions that arrive with a recognised referrer. The 35–70% of AI sessions that land in Direct are invisible to it, exactly as they’re invisible to a custom channel group.
That’s why the native channel and your custom group solve the same visible half of the problem. Neither recovers the referrer-less sessions sitting in Direct. Which brings us to the part almost no guide covers properly.
Interrogating your Direct traffic
You can’t perfectly separate AI-driven Direct sessions from genuine Direct sessions. But you can triangulate, and the signal is usually clear enough to act on. In my experience auditing analytics setups, three checks do most of the work:
- Landing page mismatch. Genuine Direct traffic lands disproportionately on the homepage and known URLs people bookmark. AI Direct traffic lands on deep, specific article and product URLs nobody types by hand. A spike in Direct sessions to a buried blog post is almost always an AI or dark-social citation.
- New vs returning. AI-referred visitors skew heavily toward new users. If your Direct traffic to a given page is 80%+ new users, it isn’t people returning to a bookmark.
- Timing against publication or citation events. Cross-reference Direct spikes with when a page started getting cited. Tools that monitor AI answer visibility can tell you when a URL entered ChatGPT or Perplexity responses; line that up against your Direct curve.
None of this is exact. It doesn’t need to be. The goal is a defensible estimate, not a false decimal place.
What to actually measure once you can track AI traffic in GA4
Volume is the boring metric. Once the plumbing works, these are the numbers that change decisions.
Conversions, not just sessions. AI-referred visitors tend to arrive later in the research journey — they’ve already had the AI summarise their options. Segment your custom channel and look at conversion rate and assisted conversions, not just traffic count. In several projects I’ve worked on, AI channels showed lower volume but meaningfully higher intent than generic organic.
Which pages get cited. Your landing-page breakdown is a live map of your AI-visible content. Pages that pull AI traffic are the ones with clear, quotable, well-structured answers. Pages that don’t are your optimisation backlog.
Source split. ChatGPT, Perplexity, Gemini and Copilot behave differently and serve different audiences. Perplexity users skew more research-heavy; Copilot leans enterprise. Knowing your split tells you which engines to prioritise in your generative engine optimization work.
Trend, not snapshot. With AI referral traffic potentially heading toward a fifth of all referral traffic, the month-over-month slope matters more than any single reading. Set up the tracking now so you have a baseline before the curve steepens.
A step-by-step framework you can run this week
Here’s the full sequence, start to finish:
- Build the custom channel group with the AI Assistants regex above, ordered above Referral.
- Enable and inspect GA4’s native AI Assistant channel as a cross-check against your custom group.
- Create a saved report or exploration filtered to AI channels, with Landing page and New/Returning as dimensions.
- Audit Direct traffic for deep-URL landing pages with high new-user rates — flag them as probable AI sessions.
- Layer in conversions by adding your key events to the AI channel report.
- Set a quarterly review to add newly launched assistants and re-check referrer behaviour, which changes often.
Run that once and you’ve moved from guessing to measuring. It’s the same discipline I bring to any SEO audit — you don’t optimise what you can’t see.
Worth noting why this matters strategically: earned editorial coverage now accounts for the overwhelming majority of AI citations — one 2026 analysis put earned media at 84% of all AI citations versus 0.3% for paid content. AI engines trust the same authoritative, well-cited content that has always won in search. So the pages showing up in your AI channel report are a direct read on where your topical authority is strongest.
Turning AI traffic data into strategy
Tracking is the start, not the end. Once you can see which content earns AI citations and traffic, the strategy writes itself: produce more of what gets cited, restructure what doesn’t, and feed the engines the clear, quotable, data-backed answers they reward.
This is where measurement and SEO consulting meet. The teams pulling ahead in 2026 aren’t the ones with the most AI traffic — they’re the ones who can see it, attribute it, and act on it while everyone else argues about whether AI search is real. Learning to track AI traffic in GA4 is the unglamorous, decisive first move. Do it before your competitors do.
Frequently Asked Questions
Does tracking AI traffic in GA4 cost anything?
No. Custom channel groups and the native AI Assistant channel are both part of standard, free GA4. The only cost is the ten minutes of setup and a quarterly review. Third-party AI-visibility monitoring tools are optional and paid, but you don’t need them to build the core tracking described here.
How accurate is AI traffic tracking in GA4?
The referred portion — sessions that arrive with a recognised AI source — is accurate and reliable. The Direct-hidden portion can only be estimated using landing-page, new-user and timing signals. Treat your reported AI channel number as a firm floor and your Direct analysis as an informed estimate of the true total.
Can I track AI traffic conversions and not just visits?
Yes, and you should. Add your key events and conversions to any report filtered by your AI channel group. Because AI-referred visitors often arrive later in the buying journey, their conversion behaviour is frequently the most interesting part of the whole dataset — sometimes more valuable than the raw session count.
How often should I update my AI channel group?
Quarterly at minimum. New AI assistants launch regularly, and existing ones change how — or whether — they pass referrer data. A channel group left untouched for a year will silently undercount as new sources appear and slip into Direct or generic Referral. Diarise the review the same way you would a technical audit.
