You track traffic from ChatGPT and other AI assistants in Google Analytics 4 with the AI Assistant channel, which Google added to the default channel group in May 2026. If you want to decide yourself which sources count, create a custom channel group with a regex on session source. Clicks from Google AI Overviews and AI Mode count as organic search traffic in GA4.
This step-by-step guide assumes a working Google Analytics 4 property in which you have at least the Editor role. Menu names follow the English interface, as Google uses them in its help pages.
What does GA4 count as AI traffic?
An AI assistant is a service such as ChatGPT, Gemini or Claude that answers questions in plain language. Google describes the AI Assistant channel as visits from sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok. In its May 2026 announcement, Google also names Claude. When GA4 recognises the referring website as an AI assistant, the visit automatically gets these values:
| Dimension | Value |
|---|---|
| Medium | ai-assistant |
| Campaign | (ai-assistant) |
| Default channel | AI Assistant |
The channel excludes Google’s own AI Overviews and AI Mode. According to the channel definitions, those clicks fall under Organic Search.
Step 1: look at the AI Assistant channel
- Go to Reports > Acquisition > Traffic acquisition.
- Leave the primary dimension set to Session default channel group.
- Find the AI Assistant row.
- Choose a period of at least a few months and compare it with the period before.
If the row isn’t there, GA4 saw no visits from recognised AI assistants in that period. Bear in mind that data from before the May 2026 launch may be classified differently, so it is better to compare two periods that both fall after it.
Step 2: see which services send the traffic
- In the same report, click the plus sign next to the primary dimension and choose Session source as the secondary dimension.
- Type
ai-assistantinto the search bar above the table, or filter on the AI Assistant row.
You now see how many sessions came in from each service. The source is usually the domain of the referring service, such as chatgpt.com, perplexity.ai or gemini.google.com. Check which values appear in your own property before drawing conclusions.
Step 3: create a custom channel group (optional)
You can’t edit the default channel group. If you want your own rules, for example because a service is missing, you make a copy.
- Go to Admin > Data display > Channel groups.
- Click Create new channel group. GA4 starts from a copy of the default group.
- Click Add new channel and give it a name, such as “AI assistants”.
- Set the condition to Session source, matches regex, and enter a regex such as:
.*(chatgpt|openai|perplexity|gemini\.google|copilot|claude\.ai).*- Click Reorder and move your new channel above Referral. GA4 assigns a visit to the first channel whose rules match. If Referral comes first, your AI visits end up there.
- Click Save group.
This is why the order matters:
How GA4 assigns a session to a channel
- VisitSomeone clicks a link to your site in ChatGPT; the session source is then, for example, chatgpt.com.
- RulesGA4 works through the channels in your group from top to bottom.
- First matchThe first channel whose conditions match gets the session.
- ReportYou see the session under that channel in Traffic acquisition and in your explorations.
According to Google’s regex guide, regular expressions in Google Analytics are a full match by default and case-sensitive, which is why there is .* on both sides. The word copilot can also catch sources other than Microsoft Copilot, so check after a few weeks which sources fall into the channel. Update the regex as new services appear; Google gives the same advice in its example for AI assistants.
A custom channel group is applied to your existing data in reports and explorations straight away. You can then select it as the primary or secondary dimension in the Traffic acquisition report.
Step 4: see what AI visitors do
Session counts mean little without behaviour. Use an exploration:
- Go to Explore and start a blank exploration. The default technique is a free-form table.
- Add Session default channel group (or your custom channel group), Session source and Landing page + query string as dimensions.
- Add Sessions, Engaged sessions and Key events as metrics.
- Drag the channel and landing page into the rows and filter on the AI channel.
You can now see which pages AI visitors land on and whether they reach an enquiry, a purchase or another key goal. A frequently used landing page gives a hint about which content the assistants use as a source. Why one page appears as a source and another does not is explained in How do AI assistants choose which websites to mention?
Check your AI channel after a few weeks
- Look at which Session source values fall into your own channel
- Remove sources that have nothing to do with AI assistants, such as another site with copilot in its name
- Add new AI services that you see turning up under Referral
- Only compare periods that both fall after May 2026
- Note which landing pages AI visitors open most often
- Check whether those visitors reach a key event, such as an enquiry or a purchase
Step 5: compare with Google Search Console
For Google’s own AI features, look in Google Search Console, not in GA4:
- Clicks and impressions from AI Overviews and AI Mode are included in the performance report under the Web search type, as Google explains in AI features and your website.
- The separate generative AI performance report shows how often links to your site were displayed in AI Overviews and AI Mode, by page, country, device and date. Google announced it on 3 June 2026, and it has been available to all websites since 31 August 2026. The report counts impressions, not visits: you find those in GA4 under Organic Search.
Which report shows what?
| Google Analytics 4 | Google Search Console | |
|---|---|---|
| What does it count? | Visits and what visitors do next | Impressions and clicks in Google Search |
| Clicks from ChatGPT, Perplexity or Claude | AI Assistant channel, or your own channel | Not visible |
| Clicks from AI Overviews and AI Mode | Under Organic Search, not separately | Performance report, Web search type |
| Impressions of your links in AI Overviews and AI Mode | Not visible | Generative AI performance report |
| Behaviour after the click | Landing page, engaged sessions, key events | Not visible |
What GA4 doesn’t show
GA4 only measures visitors who click through and whose source is known. Keep these limits in mind:
| Situation | What GA4 sees |
|---|---|
| Someone reads your name in an AI answer and later types your address | Direct traffic |
| The link runs through a redirect or shortened link that strips the source | Often direct traffic |
| The visitor uses an ad blocker | Possibly direct traffic, or nothing at all |
| A click from AI Overviews or AI Mode | Organic Search |
| An AI answer names you but nobody clicks | Nothing |
The causes of direct traffic are listed in Google’s explanation of (direct) / (none). Depending on your cookie settings, GA4 may also not record every visit in full. So the AI channel gives you a lower limit: at least this many visits came in through a recognised assistant.
How Outseen approaches this
Traffic is one source of data. Whether and how AI assistants name your business is something Outseen examines separately: we currently check AI answers by hand, with questions, environment and measurement moment agreed in advance. How to set up such a measurement yourself is explained in measuring visibility in AI answers. What else GEO involves is covered in what is GEO and on our GEO service page.
What does this mean for your business?
First check whether the AI Assistant channel appears in your Traffic acquisition report and which services sit behind it. If visits are coming in, open the landing pages: those are the pages assistants already use as a source, and so the pages where your information has to be right. If no visits come in, that does not mean you are never mentioned, because GA4 cannot see a mention without a click.
Use the figures as a trend rather than a total. Compare month on month, and set them alongside the performance report in Search Console and a manual check of the answers themselves. Only when those three sources point the same way do you have a picture you can base decisions on.