Yes, you can track ChatGPT traffic in Google Analytics 4, and it is more straightforward than most people expect, because a click from inside ChatGPT arrives on your site as an ordinary referral. The reason so many teams think they cannot see it is that they never look in the right report, and that a real portion of AI traffic quietly lands in Direct where GA4 cannot label it. So this guide does two things: it shows you exactly how to find and group ChatGPT traffic in GA4, and it is honest about the part GA4 will never show you, so you interpret the numbers correctly instead of trusting a figure that is structurally incomplete.
The short answer
To track ChatGPT traffic in GA4, open Reports, Traffic acquisition, and filter Session source / medium for chatgpt.com / referral and chat.openai.com / referral, which fall under Referral in the default channel grouping (Google Analytics Help). For ongoing tracking, build a custom channel group named AI Traffic that matches the AI referral domains with a regex (Google Analytics Help). The essential caveat: GA4 only captures AI visits that pass a referrer, so referrer-less visits land in Direct (Google Analytics Help). Every GA4 number for AI traffic is a floor, not a ceiling.
How ChatGPT traffic appears in GA4
Start with the mechanism, because it explains everything else. GA4 attributes a session by its referrer, the site the visit came from. When a user clicks a cited link inside ChatGPT and their browser passes the referrer, GA4 records the source as chatgpt.com (or chat.openai.com for some links) and the medium as referral. In GA4’s default channel group, that traffic is bucketed under Referral alongside every other third-party link (Google Analytics Help). So ChatGPT traffic is not invisible; it is sitting in your Referral channel, mixed in with everything else, waiting for you to isolate it. The whole task is separating it out and accepting what the referrer cannot tell you.
The quick check: Traffic acquisition
The fastest way to see it right now takes under a minute. Go to Reports, then Acquisition, then Traffic acquisition. Change or add the dimension to Session source / medium, then use the search box to filter for chatgpt. You will see rows like chatgpt.com / referral and possibly chat.openai.com / referral, with sessions, engagement, and any conversions for that traffic. That single view already tells you whether ChatGPT is sending measurable visits and how engaged they are. It is the right first step before building anything, because it confirms the traffic exists and shows you the exact source strings you will need for the more durable setup.
Build an AI Traffic channel group
The quick check is manual; the durable solution is a custom channel group. In Admin, under Data display, open Channel groups and create a new one. Add a channel named AI Traffic and set a condition where Source matches a regex covering the AI domains, something like chatgpt.com|openai.com|perplexity.ai|gemini.google.com|copilot.microsoft.com|claude.ai (matches regex). Order it above the default Referral channel so AI visits are claimed by AI Traffic rather than lumped into generic Referral. Custom channel groups are exactly the supported tool for this (Google Analytics Help), and once built, every report that offers the channel dimension can show AI traffic as its own line, which is far more useful than digging through Referral each time.
Which AI domains to include
Do not track only ChatGPT if your goal is AI traffic, because the same technique covers every assistant. Include the referral domains for the major engines: ChatGPT (chatgpt.com, chat.openai.com), Perplexity (perplexity.ai), Google’s Gemini (gemini.google.com), Microsoft Copilot (copilot.microsoft.com), and Claude (claude.ai). Grouping them into one AI Traffic channel lets you compare engines and see the whole AI-referral picture, and it future-proofs the setup as usage shifts between assistants. You can also keep per-engine channels if you want the breakdown. The point is to define the set deliberately from real referral strings you have verified in your own data, not to guess, since the exact hostnames matter for the regex to match.
The big caveat: traffic hiding in Direct
Here is the part most guides underplay. GA4 can only attribute a visit if a referrer is present, and a meaningful share of AI traffic arrives with none, for example clicks from within a mobile app or links where the referrer is stripped. GA4 files those as Direct with source (direct) and medium (none), which by definition means no referral information was available (Google Analytics Help). That traffic is real ChatGPT-driven visiting that GA4 simply cannot label, so it disappears into Direct. This is the same effect explored in do clicks from Perplexity AI show as direct traffic: AI referral tracking always undercounts, and the size of the gap is not directly measurable.
Where AI traffic shows up in GA4
This table summarises what you can and cannot see.
| Visit type | GA4 attribution | Trackable? |
|---|---|---|
| ChatGPT web click passing referrer | chatgpt.com / referral | Yes, in Referral / AI Traffic |
| Older ChatGPT link | chat.openai.com / referral | Yes, in Referral / AI Traffic |
| Perplexity, Gemini, Copilot, Claude clicks | their domain / referral | Yes, if added to the regex |
| App click or stripped referrer | (direct) / (none) | No, hidden in Direct |
| Zero-click AI answer (no visit) | not recorded | No, never a session |
Why GA4 shows a floor, not a ceiling
Combine the Direct gap with a second reality and the interpretation becomes clear. Much of AI’s influence produces no click at all: when an answer engine summarises your content in place, the user is informed without visiting, and Pew found people click a result on just 8% of pages with an AI summary versus 15% without (Pew Research). So GA4’s AI Traffic number reflects only the referral-passing clicks, missing both the Direct-hidden clicks and the entirely clickless influence. Treat the figure as a conservative floor of AI impact, never the full story, which is why measuring AI value needs more than click counts, as covered in how to measure SEO ROI when generative AI gives zero-click answers.
Go deeper with an Exploration
Once the channel group exists, use Explore for real analysis. Create a blank Exploration, set a long date range, add Session source / medium or your AI Traffic channel as a dimension, and add Landing page plus query string as a second dimension. Bring in metrics like Sessions, Engaged sessions, Engagement rate, and your key events or conversions. Now you can see which pages ChatGPT sends people to, how engaged those visitors are, and whether they convert. That page-level view is where tracking becomes actionable: it tells you which of your content is actually being cited and clicked, so you can do more of what works rather than guessing.
What to measure once you can see it
With the setup done, focus on the metrics that matter rather than vanity sessions. Watch which landing pages earn AI referrals, since those are your cited, working pages. Compare engagement and conversion of AI traffic against other channels for your own site, because the quality can differ. Track the trend over time as your citations grow or fade. And connect drops to cause, if AI referrals to a page fall, it may have lost a citation, the diagnostic in why is ChatGPT referral traffic dropping. Measuring the right things turns a channel report into a feedback loop for your AI-visibility work.
Do not trust the circulating vendor stats
A word of caution on numbers you will see quoted. Various blogs cite precise figures for AI referral share, year-over-year growth, and conversion lift, and these get repeated as if settled. They are not authoritative, they vary wildly by source and sample, and your own site’s reality is the only number that matters for your decisions. Use GA4 to measure your actual AI traffic rather than importing someone else’s percentages. The honest position is that AI referral traffic is real and worth tracking, that it is growing for many sites, and that the exact magnitudes are unsettled, so measure yours and treat external percentages as loose context at best.
How this connects to citations
Tracking closes a loop that starts with citations. An AI referral click only happens when an engine cited or drew on your page and the user chose to click through, and models cite the sources that best and most currently answer the query, as Profound’s citation-pattern analysis shows (Profound). So your AI Traffic channel is, in effect, a downstream signal of which cited content converts attention into visits. That is why GA4 tracking pairs naturally with citation monitoring: citations tell you where you appear in answers, and GA4 tells you which of those appearances actually pull people to your site. Together they give the fuller picture that either alone misses.
A worked example
A team suspects ChatGPT is sending traffic but cannot find it. They open Traffic acquisition, filter Session source / medium for chatgpt, and immediately see chatgpt.com / referral sessions concentrated on three help articles. They build an AI Traffic channel group covering ChatGPT, Perplexity, Gemini, Copilot, and Claude, then an Exploration by landing page. They learn those three articles are cited and clicked, engagement is strong, and a handful convert. They also note Direct rose in the same period, likely masking more AI traffic. So they treat the GA4 number as a floor, double down on the cited articles, and add citation monitoring. They went from AI traffic is invisible to a working, honest measurement setup in an afternoon.
Common mistakes
The first mistake is looking in the wrong report and concluding AI traffic is untrackable, when it sits in Traffic acquisition under Referral. The second is tracking only chatgpt.com and missing the other assistants and the chat.openai.com variant. The third is treating GA4’s AI number as complete, ignoring the Direct-hidden clicks. The fourth is importing vendor percentages instead of measuring your own. The fifth is counting sessions without checking landing pages and conversions, which is where the insight lives. Avoid these and you get a setup that is both accurate about what it shows and honest about what it cannot.
The bottom line
How do you track ChatGPT traffic in GA4? Find it in Traffic acquisition as chatgpt.com / referral, build a custom AI Traffic channel group covering the major assistants with a regex, and analyse it by landing page and conversion in an Exploration. Then interpret it correctly: GA4 only sees referrer-passing visits, so app and stripped-referrer clicks hide in Direct and clickless AI influence never appears at all, making every number a floor. Measure your own traffic rather than trusting circulating percentages, and pair GA4 with citation monitoring so you see both where you are cited and which citations actually drive visits.
Frequently asked questions
How do I track ChatGPT traffic in GA4?
Open Reports, Traffic acquisition, and filter the Session source / medium dimension for chatgpt.com / referral and chat.openai.com / referral. For ongoing tracking, build a custom channel group called AI Traffic that matches the AI referral domains with a regex, so ChatGPT and other AI referrals report together. Use an Exploration to break it down by landing page and conversions. Remember GA4 only sees AI visits that pass a referrer.
Why does some ChatGPT traffic show up as Direct in GA4?
Because GA4 attributes traffic by the referrer, and when a visit arrives with no referrer information, such as clicks from an app or a stripped referrer, GA4 files it as Direct with no source. So a portion of real ChatGPT traffic lands in Direct and goes uncredited. This is why GA4 reports a floor for AI traffic, not a ceiling; your true AI-driven visits are higher than any report shows.
What source and medium does ChatGPT traffic use in GA4?
Referral-passing ChatGPT visits typically appear as chatgpt.com / referral, and older or app links may appear as chat.openai.com / referral. In GA4’s default channel grouping these fall under Referral. Building a custom AI Traffic channel group lets you separate them from ordinary referral traffic and group them with Perplexity, Gemini, Copilot, and Claude referrals.
Can I fully measure how much traffic ChatGPT sends me?
Not exactly. GA4 captures the AI visits that pass a referrer, which is a reliable floor, but some AI traffic hides in Direct and cannot be attributed, so no GA4 report gives the complete number. Treat GA4’s AI figures as a conservative minimum, pair them with citation monitoring to see where you are cited, and do not rely on circulating vendor percentages as fact.
