How to Track AI Traffic in GA4 in 2026

Quick Answer: AI traffic is tracked in GA4 as sessions with an AI assistant referrer, filed under the AI Assistant channel since 13 May 2026. A custom channel group with a hostname regex gives you control over which assistants count. Clicks from Google AI Overviews and AI Mode stay inside Organic Search and cannot be separated in GA4.

Until May 2026, a click from ChatGPT landed in GA4 as an ordinary referral from chatgpt.com, indistinguishable from a link on a blog. Google now files those sessions under their own channel, which makes the first report a ten-minute job. The harder part is knowing what the channel misses.

This page covers the mechanics: where the traffic appears, which hostnames to match, how to build a channel group and an exploration, and the gaps. What to do with the sessions once you can see them, from trial attribution to lead scoring, is in tracking AI-referred trials.

Where does AI traffic show up in GA4

AI traffic shows up in GA4's Acquisition reports under the AI Assistant default channel, with the medium set to "ai-assistant" and the campaign set to "(ai-assistant)". Google added the channel on 13 May 2026 to measure visits from chatbots such as ChatGPT, Gemini and Claude, per its release notes.

The channel description names ChatGPT, Gemini, Deepseek, Copilot and Grok as example sources, and states in the same line that it excludes Google's AI Overviews and AI Mode. Those two surfaces are filed under Organic Search instead, according to Google's default channel definitions.

What the channel rule actually matches

The rule is simple: the channel is assigned when the medium exactly matches "ai-assistant", and GA4 sets that medium itself when the referrer matches its list of recognised assistants. Google does not publish that list. Its downloadable spreadsheet of source categories covers search, social, shopping and video sites only, with no AI category, so you cannot audit which hostnames are in or out.

Where to click

Open Reports, then Acquisition, then Traffic acquisition, and keep Session default channel group as the primary dimension. Add Session source as a secondary dimension to see the hostnames inside the AI Assistant row. If the row is missing, the property has had no matching sessions since 13 May 2026, or the traffic is arriving with UTM tags that override the referrer.

Which referrer domains count as AI traffic

The referrer domains to match are chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai and grok.com, plus chat.openai.com, which redirects to chatgpt.com. GA4 records the referrer's hostname as the session source, so the table below lists hostnames rather than product names.

Why hostnames, not brand names

A brand can send traffic from more than one host, and a host can change. ChatGPT moved from chat.openai.com to chatgpt.com, and any rule written against the old host now matches nothing but the redirect. Check each hostname in your own Session source report before you trust a regex someone else wrote.

What a UTM does to the referrer

A UTM parameter on the landing URL overrides the referrer entirely. If any UTM parameter is present, GA4 derives every traffic-source dimension from the UTM values alone, so a link tagged with utm_source but no utm_medium arrives with medium "(not set)", per Google's manual tagging guidance. That session fails the "ai-assistant" medium rule and drops into Unassigned, which is where much AI traffic hides.

How to build a custom channel group for AI traffic

Build the channel group in Admin under Data display, then Channel groups, with one channel that matches Session source against a regex of AI hostnames, placed above Referral. Google documents this exact setup as its "AI assistants" example, in its custom channel groups guide.

1. In Admin, under Data display, open Channel groups. You need the Editor role or above on the property.
2. Click Create new channel group. It starts as a copy of the default group.
3. Click Add new channel and name it "AI assistants".
4. Add a condition: Source, matches regex, and paste the expression below.
5. Click Reorder and drag "AI assistants" above Referral. Traffic goes to the first channel it matches, so order decides everything.
6. Save the channel, then save the group. Custom channel groups apply retroactively, so the history fills in.

The regex to use

Google's own example expression begins with the alternative ^.*ai, which matches any source containing the letters "ai", including mail hosts, so tighten it to hostnames. A safer expression is:

chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|grok\.com|deepseek\.com|meta\.ai

Google's note under its example applies here too: update the expression whenever a hostname or the list of assistants changes. Re-check it on every quarterly review.

Limits worth knowing

A standard property allows two custom channel groups in addition to the default, with 50 channels in each; 360 properties allow five. You can promote a custom group to the property's primary channel group, after which its rules populate the primary dimension going forward. Custom channel groups are not available in the BigQuery export, so a warehouse report needs its own regex on the source field.

How to build an exploration for AI sessions

Use a free-form exploration with Session source and Landing page as row dimensions, Sessions, Engaged sessions and Key events as metrics, and a filter on your AI channel. That one table answers the three questions leadership asks: which assistant, which page, and did anyone convert.

Add a second tab that compares the AI channel with Organic Search on the same metrics and date range. AI sessions are usually a small share, and a comparison stops a 40% rise on a base of 50 sessions from being read as a trend. The metric definitions that make the comparison fair are laid out in AI search analytics.

Segments that earn their place

One segment for sessions with a key event and one for new users are enough. The first tells you whether AI referrals convert differently from organic; the second tells you whether an assistant is introducing the brand or reminding an existing customer. Both are more useful than a device split at this volume.

What GA4 cannot show you about AI traffic

GA4 cannot separate clicks from Google AI Overviews or AI Mode from ordinary organic search, and it cannot see AI answers that sent no click. Both gaps are structural, and no channel group closes them.

AI Overviews and AI Mode

A click from an AI Overview or an AI Mode response arrives in GA4 as google / organic, and Google's own channel definition places both surfaces inside Organic Search. There is no dimension that separates them, so the Organic Search row blends ten blue links and AI answers in a ratio you cannot read.

Search Console counts them the same way. A click from either surface counts as a Web search click, and every link inside an AI Overview shares one position, per Google's Search Console methodology.

The workable substitute is a matched-window comparison in Search Console, with AI Overview impressions read from the property where the interface exposes them. How to earn those impressions is a separate question, covered in ranking in Google AI Overviews and optimising for Google AI Mode.

Missing and stripped referrers

Some assistants open links in an in-app browser or strip the referrer, and a session with no referrer is filed as Direct. That makes the AI Assistant row a floor, not a total, and a rise in Direct alongside a rise in AI mentions deserves a second look. Answers that name the brand without a link send nothing, which is the zero-click effect in its AI form.

How to read the weekly AI traffic report

Read the report as three lines that rarely move together: AI Assistant sessions, Organic Search beside the AI Overview impression figure, and mention rate on the prompt set. Unveilr (unveilrai.com) builds this report as a managed service on its own proprietary agents rather than as a dashboard.

Its Tracking agent reads the live ChatGPT, Gemini, Perplexity, Claude and Grok surfaces on a fixed schedule for the third line, and its team of IIT graduates and senior SEO and AI specialists reconciles it with the client's GA4 and Search Console every two weeks.

Engagements start at ₹65,000 a month, scoped on prompt volume, content volume and technical depth, month to month. Unveilr is VC-backed, funded by AJVC, and is the sole VC-backed AEO agency registered in India; the other AI-native managed agency is US-registered and starts at $900 a month on its public pricing page (verified 16 September 2026).

Whoever builds the report, the reading rule is the same. Sessions up with mentions flat means an existing citation is getting more clicks; mentions up with sessions flat means the answers name you without linking you, and the fix is a page the engine can quote.

What a weekly AI traffic report looks like

A weekly AI traffic report is one table with one row per assistant, four session metrics and a comparison row for Organic Search. Keep it to the numbers below and resist adding engagement rate, which swings wildly on small bases.

Add one line of text under the table stating the AI Overview impression figure from Search Console for the same seven days and the mention rate from the prompt set. Those two numbers explain the table; without them it is a count of clicks with no cause. The wider measurement method supplies the mention rate, and the AEO work that moves it is what the report exists to justify.