How to Identify AI Referral Traffic in GA4

Yes, you can isolate the AI referral traffic GA4 buries inside its default Referral, Direct and Unassigned channels. The setup takes about an hour: build a regex referrer list covering chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai, wrap it in a custom channel group called AI Assistants, then validate everything with a free-form exploration keyed on session source. You end up with a durable line item showing which assistants send visitors, which pages they land on and what those sessions actually convert to.

Why AI Referrals Vanish Inside GA4's Default Reports

Open the default traffic acquisition report and look for your ChatGPT sessions. You will not find them, because GA4's default grouping has no AI channel. The sessions exist. They are just shredded across three buckets, so the trend never becomes a line you can point at in a meeting. A trend nobody can point at is a trend nobody budgets for.

Picture the moment that makes this sting. A strong citation week sends perplexity.ai visitors to a comparison page. Today that spike disappears inside Referral, buried beside newsletters and partner links. Tomorrow, with one channel group in place, it becomes the first thing your CMO sees.

Where GA4 currently buckets ChatGPT, Perplexity and Gemini sessions

When an assistant passes a clean referrer, GA4 files the session under Referral with every other website that ever linked to you. When the referrer gets stripped, which happens constantly in mobile apps and in-app browsers, the session lands in Direct. When GA4 cannot make sense of the source at all, it goes to Unassigned, the junk drawer of analytics.

Referral vs Direct vs Unassigned: the three hiding places

The messiness is structural, so the fix should be structural too. One regex list, one custom channel group, one exploration to validate. That is the whole project. The payoff is a durable piece of AI visibility measurement your team can watch week over week instead of reconstructing from memory every quarter.

Which AI Assistants Actually Send Referral Traffic?

Before you touch GA4 settings, know what you are matching. The list of assistants that send meaningful clicks is short enough to memorize, and their referrer domains are stable enough to pattern-match.

The referrer domain list to start from

Notice how differently these engines behave. Perplexity cites and links inline almost compulsively, so perplexity referral traffic tends to arrive in steady drips tied directly to citations. ChatGPT links more selectively, which means chatgpt referral traffic often looks lumpier: quiet for weeks, then a burst when a page gets pulled into answers for a popular question. Add edge cases like you.com and Brave's Leo if your audience skews technical.

Why AI Overviews need a different method entirely

Google's AI surfaces are a separate problem, and the boundary should stay crisp. Clicks from AI Overviews arrive as google / organic, indistinguishable from classic search inside GA4, so no referrer regex will ever catch them. We covered that method in How to Track AI Overviews Traffic in Search Console. Keep the two workstreams separate and both stay defensible. One honest caveat before the build: this setup captures clicked referrals only, and plenty of AI influence never produces a click at all. More on that gap below.

The AI Referral Traffic GA4 Setup: Regex Lists and Channel Groups

Now the build. If regex makes you nervous, relax. This one is about as gentle as regex gets, and you can paste it verbatim.

The regex referrer list, line by line

The pattern matches session source against every assistant domain worth catching:

.*chatgpt.*|.*openai.*|.*perplexity.*|.*gemini.*|.*copilot.*|.*claude.*

Reading it token by token: .* means anything before or after, each domain keyword sits between two wildcards, and the pipe character | means or. The whole expression translates to any source containing chatgpt, openai, perplexity and the rest. The wildcards matter because they catch subdomain variants. chat.openai.com and chatgpt.com both match without you enumerating either.

Building the custom channel group in five clicks

1. Go to Admin, then Data settings, then Channel groups.
2. Copy the default channel group rather than editing it - keep the original intact as a fallback.
3. Create a new channel named AI Assistants.
4. Set the condition to "session source matches regex" and paste the pattern.
5. Save, then open a traffic acquisition report and switch the primary dimension to your new group.

Ordering rules so 'AI Assistants' wins before 'Referral'

One detail decides whether this works: evaluation order. GA4 walks the channel list top to bottom and assigns each session to the first rule it matches. If Referral sits above AI Assistants, Referral claims every chatgpt.com session before your rule gets a look. Drag AI Assistants above Referral, and the channel starts winning the sessions it deserves. Then put a recurring block on your calendar, because assistants change domains without announcement. chat.openai.com quietly becoming chatgpt.com is the canonical example.

How Do You Build the Exploration Report That Proves the Trend?

The channel group makes the traffic visible. The exploration makes it persuasive. Explorations also earn their keep in ways standard reports cannot: you can segment periods from before the group existed, split by device to see how much mobile stripping costs you, and inspect anomalies one referrer at a time.

The exact dimensions and metrics to drop in

Build a free-form exploration with session source / medium as the dimension, filtered by the same regex you used in the channel group. For metrics, pull in sessions, engaged sessions, key events and session conversion rate. Add a week-over-week line chart as the primary visualization so the trend reads at a glance, the way TechTarget's Google Analytics 4 reference recommends structuring any trend-first report.

Landing page analysis: which content earns AI clicks

Then add landing page as a breakdown dimension, because this is where the strategy lives. Knowing that Perplexity sent 80 sessions is trivia. Knowing that 60 of them landed on your pricing comparison page is a content brief. Save the exploration and share it with the team so everyone works from one canonical view of how to track llm traffic in analytics, rather than five private variants that never quite agree.

From Sessions to a Real AI Visibility Line Item

Once the exploration confirms the data is clean, promote the channel from curiosity to reporting artifact. Put AI Assistants in the monthly deck beside Organic Search and Paid, with the same three columns everything else gets: sessions, engagement rate, conversions. The moment it sits in the same table as your other channels, people stop treating it as a novelty and start asking useful questions.

Pairing referral data with citation share

Referral traffic is the lagging indicator. The leading one is citation share: how often assistants mention and cite you across the prompts your buyers actually ask. Read the two together and a rhythm emerges. Citations rise first, clicks follow weeks later. When a page suddenly draws Perplexity sessions, check which prompts cite it, figure out what about the format earned the citation, and double down on that pattern across your other pages.

What to report weekly vs monthly

Weekly, watch the exploration for movement: new landing pages appearing, a single assistant surging, anything that changed. Monthly, report the channel-level rollup in the same frame as everything else. Consistency matters more than precision here. A metric measured the same way every month builds trust even when the absolute numbers are modest.

What This Setup Can and Cannot Tell You

Being honest about the limits makes the whole thing more credible in the room where budgets get decided, so state them plainly.

The dark traffic problem, stated plainly

Stripped referrers mean some AI sessions will always masquerade as Direct. Your AI Assistants channel is therefore a floor rather than a census, and that is genuinely fine. A consistent floor beats an inflated guess every time. If Direct spikes in lockstep with citation gains, you have a decent hint about where the missing sessions went.

Pairing GA4 with prompt-level monitoring

The complete picture takes two instruments. Run an AI visibility audit for prompt-level presence: where you appear, where competitors appear, where nobody does. Keep GA4 for click-level proof. The honest gap between the two is not a flaw in either measurement; it is the accurate size of AI search traffic attribution today. For the bigger question of whether all this compounds into pipeline, we took a hard look at whether AI visibility drives traffic and revenue.

Automating the Weekly Read So Nobody Babysits GA4

The manual version works. It also decays, because the report that requires someone to remember it gets forgotten by week six. Anyone who has tried to sustain a manual reporting workflow, the kind Zapier's GA4 reporting guides exist to patch over, knows exactly how that story ends.

What an agent-run weekly digest looks like

This is the gap AstroFabric closes. Its AI visibility agent tracks assistant citations upstream while the performance agent reads the downstream traffic picture. The week-over-week math runs in a code sandbox as exact computation, so the delta in your Monday digest is precise rather than eyeballed from a chart. The digest lands wherever your team already lives: Slack, Telegram or email. That is what makes the habit stick.

Keeping humans on the approval gate

Any write action the agents propose sits behind explicit approval, so nothing changes in your stack without a human saying yes. The promise stays modest and concrete: an hour of setup today, a defensible AI channel in your next reporting cycle, and a system that keeps reading the numbers after everyone else has moved on to the next fire.

See Your AI Channel This Week

Build the regex, save the channel group, validate with one exploration, then let the agents keep the watch. Start with AstroFabric and turn a hidden trend into the line item your next reporting cycle deserves.

Frequently asked questions

Sources

• Zapier's guide to GA4 reporting workflows
• TechTarget's Google Analytics 4 reference