How to See AI Traffic in GA4: Step by Step

Short answer: Yes, GA4 can show you AI traffic. Since May 2026 there is a native AI Assistant default channel group that automatically groups visits from recognised AI chatbots such as ChatGPT, Gemini and Claude. To view it, go to Reports > Acquisition > Traffic acquisition and set the primary dimension to Session default channel group, then look for AI Assistant. For fuller coverage you can also build a custom channel group with a regular expression that matches AI hostnames, which lets you add engines yourself. Both methods only ever see AI referrals, which are humans clicking a link out of an AI answer. They never see AI crawlers, and both systematically undercount because AI platforms frequently strip the referrer, dropping the visit into Direct.

Does GA4 have a built-in AI traffic channel?

Yes. In May 2026 Google added an AI Assistant entry to GA4's default channel groups, which Search Engine Journal reported began rolling out on 13 May 2026. When GA4 recognises a session as coming from an AI chatbot, it automatically stamps the medium as ai-assistant and the campaign as (ai-assistant). Google named ChatGPT, Gemini and Claude as examples of the platforms it recognises. Property owners do not need to configure anything for this to start working.

To see it, follow this path:

1. Open Reports > Acquisition > Traffic acquisition.
2. Set the primary dimension to Session default channel group.
3. Scan the table for the AI Assistant row.

Because GA4 also writes a specific medium and campaign, you can slice the Google-recognised subset even more tightly. Add Session medium as a secondary dimension and filter where it exactly matches ai-assistant. That gives you a clean cut of only the traffic Google is confident came from an AI assistant.

Two limitations matter before you trust this channel. First, it is forward-only and not retroactive. Traffic that arrived before the rollout keeps whatever classification it originally had, so a visit from ChatGPT in April 2026 still sits under Referral or Direct, not AI Assistant. The rollout was also gradual, reaching broad availability across properties in early June 2026, so if you do not see the channel yet it may not have reached your account.

Second, the native channel only recognises a subset of AI platforms. Google's current Default Channel Group definitions list engines including ChatGPT, Gemini, DeepSeek, Copilot and Grok, but it is a moving target and it is not a complete picture. That is where the custom method comes in.

How do I build a custom AI channel that catches more engines?

The do-it-yourself method predates the native channel and still gives you control over exactly which hostnames count. If an engine such as Perplexity, You.com or Meta AI is not covered by the native channel yet, a custom group lets you add it yourself and keep the pattern current as new platforms launch.

Here is the full path:

1. Go to Admin > Data display > Channel groups.
2. Click Create new channel group and give it a name.
3. Add a new channel, name it something like AI Traffic.
4. Set the condition to Source matches regex and paste in the AI hostname regular expression below.
5. Drag this channel above Referral so it takes precedence. Channel matching runs top-down and the first match wins, so if AI Traffic sits below Referral, these visits get swallowed by Referral first and you never see them.
6. Save.

A workable regex, matched on the Source dimension (not Medium), is:

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

Escape the dots, as shown, so they match literally rather than as regex wildcards. Like the native channel, a custom channel group applies going forward, not retroactively, so it will only reclassify traffic from the moment you save it. Keep the pattern maintained: new AI engines appear regularly, and each one you leave out is traffic that quietly falls back into Referral or Direct.

A quick check without any configuration

If you want to spot AI referrals in the next two minutes without touching Admin, open Reports > Acquisition > Traffic acquisition, add Session source / medium as the dimension, and type chatgpt, perplexity, gemini or claude into the search box. The referring sessions show up immediately. It is not a permanent grouping, but it is the fastest way to confirm AI traffic exists before you invest in setup.

Isolating the value in Explore

Once you can identify AI sessions, the interesting question is what they do. In Explore, build a segment where Source contains any of your AI hostnames, then apply it to landing pages, engagement rate, conversions and revenue. This is where AI traffic earns its keep, and we will come back to why in the value section below.

What are the main referrer hostnames for AI engines in GA4?

These are the source hostnames worth watching. The table pairs each engine with the hostname you will see in GA4 and whether the native AI Assistant channel is likely to recognise it. Google recognition changes over time, so verify against the live Default Channel Group definitions and add anything missing via the custom regex.

Match these on the Source dimension. The Google-recognised subset also carries the ai-assistant medium, so if you only want the platforms Google is confident about, filter on Session medium exactly matches ai-assistant. If you want the widest net, use the Source regex and maintain it as new engines launch.

Why does my AI traffic look lower than expected in GA4?

Because GA4 only counts AI traffic when the AI platform passes a referrer header, and AI platforms frequently do not. This is the single most important thing to understand about AI measurement in GA4.

When someone clicks a link inside a normal web browser, the browser tells your site where they came from, and GA4 reads that referrer to classify the session. But a large share of AI usage happens in places that send no referrer at all: in-app browsers inside the ChatGPT or Perplexity mobile apps, native mobile apps, and links that people copy out of an AI answer and paste into a fresh tab. When the referrer is stripped, GA4 has nothing to go on, so the session lands in Direct (recorded as (direct) / (none)).

The consequence is that both the native AI Assistant channel and any custom regex channel systematically undercount real AI referrals. Treat every AI figure GA4 gives you as a floor, not a true total. The real number is higher, you just cannot see how much higher from GA4 alone.

There is a second, deeper reason your AI numbers look small, and it is a category error rather than a measurement gap. AI referrals and AI crawlers are different things, and GA4 can only ever see referrals.

Referrals versus crawlers

An AI referral is a human clicking through from an AI answer to your page. An AI crawler is a bot fetching your pages so the AI platform can train on them, index them or ground an answer. GA4 runs on JavaScript that fires when a page renders in a real browser. Crawlers such as GPTBot, ClaudeBot, PerplexityBot and Google-Extended do not execute that JavaScript, so they never appear in GA4 at all. This is by design, not a bug.

To see which AI crawlers are hitting you, you have to read server access logs or CDN logs (Cloudflare, Fastly and similar), because that is the only place the requests are recorded. And when you do, verify the visitor by IP address, not user-agent string alone. User-agents are trivially spoofed, so anyone can claim to be GPTBot. The major providers publish official IP ranges you can check against:

• OpenAI: GPTBot (training), OAI-SearchBot (ChatGPT search citations) and ChatGPT-User (fetches a page for a live user request), with published IP ranges documented in OpenAI's crawler documentation.
• Anthropic: ClaudeBot (training), Claude-User (answering a user question) and Claude-SearchBot (search quality), all honouring robots.txt, per Anthropic's help centre.
• Perplexity: PerplexityBot (indexing, respects robots.txt) and Perplexity-User (user-initiated), documented in Perplexity's crawler docs.

One naming caveat when parsing logs: OAI-SearchBot's user-agent embeds a full browser token before the OAI-SearchBot segment, so match the substring OAI-SearchBot rather than the leading Mozilla portion.

Why the crawler gap is the real story

The reason crawlers matter economically is that AI platforms take vastly more than they send back. Cloudflare's network data measured the crawl-to-refer ratio across the first half of 2025: for every human that Anthropic's crawlers referred, those crawlers had visited tens of thousands of pages. The Anthropic ratio was 286,930 to 1 in January 2025, falling to 38,065 to 1 by July 2025. OpenAI moved from 1,217 to 1 down to 1,091 to 1 over the same period, while Google went the other way, from 3.8 to 1 up to 5.4 to 1. This is the crawl-to-click gap, and it is the reason to measure crawlers and referrals as two separate things rather than one blurred number.

Can I separate Google AI Overview clicks from normal search in GA4?

Not cleanly, and this is a genuine blind spot you should know about before you go looking. AI Overviews are the AI summaries that appear at the top of a Google search results page. When a user clicks a normal blue link on a results page that happened to show an AI Overview, that click arrives in GA4 as ordinary Organic Search. GA4 has no signal that an AI summary was even present, so you cannot isolate those sessions inside GA4 on their own.

What you can do is estimate the damage by pairing two sources. Pew Research Center (July 2025) studied 900 US adults across 68,879 Google searches in March 2025. When an AI Overview was present, users clicked a traditional search result in just 8% of visits, compared with 15% when there was no AI summary. They clicked a link inside the AI summary itself in only 1% of visits, and they ended their browsing session on 26% of pages with an AI summary versus 16% without. That is the click decoupling in hard numbers, and it explains a drop in organic clicks that no GA4 channel will attribute to AI on its own.

To triangulate it, use Google Search Console alongside GA4. Google launched a Generative AI performance report in June 2026 that shows impressions from AI surfaces broken down by page, country, device and date. The catch is severe: it has no clicks, no click-through rate, no average position and no query-level data. On its own it cannot tell you what AI is costing you. But if you pair GSC's AI-surface impressions with GA4's organic clicks to the same pages, you can estimate the click gap, which is the difference between how often your pages appear in AI answers and how rarely those appearances turn into visits.

Do not confuse the two Google AI controls

While you are here, avoid a common and costly mistake. There are two separate Google settings and they govern different things:

• Google-Extended is a robots.txt control for Gemini model training and grounding only. Per Google's own documentation, it "does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search," and it does not govern AI Overviews or AI Mode.
• The AI Overviews and AI Mode opt-out is a distinct Search Console AI features control.

Neither of these is a tracking tool, and neither the Search opt-out nor Google-Extended covers content summarised inside the standalone Gemini app. If you block Google-Extended expecting to disappear from AI Overviews, you will be disappointed, and you may cut off Gemini grounding for no measurement benefit at all.

Why bother isolating AI traffic at all?

Because it is small in volume but high in value, and leaving it buried in Referral or Direct hides your best-converting audience. Ahrefs found that over one 30-day window AI search made up just 0.5% of its traffic but drove 12.1% of signups, a conversion rate the company put at roughly 23 times that of traditional organic search. Semrush's roundup makes a related point on value, estimating the average AI search visitor as worth about 4.4 times a traditional organic visitor. For a publisher, a visitor who arrives having already read an AI summary of your topic tends to be further down the funnel and more engaged. If those sessions sit anonymously inside Direct, you cannot see that they are among your most valuable visitors, and you cannot make the case for the content that earns AI citations.

The practical takeaway is a workflow, not a single report. Turn on the native channel, extend it with a custom regex, read your server logs for crawlers, and pair GSC AI impressions with GA4 organic clicks to estimate the AI Overview gap. Then accept that the GA4 number is a floor and manage to the trend rather than the absolute.

If wiring all of that together by hand sounds like a lot, that is exactly the problem Ramprt was built to solve. It reads your existing GA4 read-only, with no new tracking script, and gives you an AI tab that shows AI referrals by engine, which AI crawlers are taking your content, how often AI answers are built on your pages, what AI-referred traffic earns against your site average, and the AI Overview click decoupling. You can see the whole thing on the free demo at ramprt.io/demo before connecting anything.

For the wider context on this shift, see our related guides on what AI is doing to your search traffic and how to track AI crawlers in your server logs.

Frequently asked questions

Is the GA4 AI Assistant channel retroactive?

No. It is forward-only. Traffic that arrived before the 13 May 2026 rollout keeps its original classification as Referral, Direct or whatever custom grouping applied. You cannot backfill historical AI traffic into the channel.

Should I match AI sources on Source or Medium?

Match on Source for the custom regex, because that is where the hostname lives. Use Medium only for the Google-recognised subset, where you can filter on Session medium exactly matching ai-assistant.

Can GA4 show me GPTBot or ClaudeBot activity?

No. Those are crawlers, and they do not execute GA4's JavaScript, so they never appear in GA4. To see crawler activity you must read server or CDN logs and verify the requests by IP against the providers' published IP ranges.

Why does so much AI traffic show up as Direct?

Because AI in-app browsers, native mobile apps and copy-pasted links often send no referrer header. With nothing to read, GA4 files the session under Direct. This means every AI figure in GA4 is a floor, and the true total is higher.

Does blocking Google-Extended remove me from AI Overviews?

No. Google-Extended only controls Gemini training and grounding. It does not affect Google Search inclusion, ranking, AI Overviews or AI Mode. The AI Overviews opt-out is a separate Search Console AI features control.