ChatGPT traffic in GA4: the AI Assistant channel and how to read it

More and more people ask ChatGPT and other AI chats for recommended companies, and some of them click a link to a website. The question is how many such people actually arrive and whether they later leave an inquiry. In the past the answer was hard to get, because visits from AI assistants got lost in other channels.

Since 13 May 2026, GA4 has had a default AI Assistant channel. In this guide I will show where to find it, what it won’t show and how to work out whether AI traffic matters for your business. I describe the situation as of 15 October 2026, and for details that may change I note that they are worth checking in GA4 Help.

What changed in GA4

GA4 groups visits into channels, for example Organic Search, Direct, Referral or Paid Search. On 13 May 2026 Google added a default AI Assistant channel to this list (described as “Asystent AI” in the Polish-language GA4 Help). It receives visits whose referrer is on the list of AI assistants: GA4 gives them the medium “ai-assistant” and the campaign “(ai-assistant)”. The channel works without configuration, but it has no historical data: Google rolled it out gradually, and the channel counts visits from the moment it appeared in your account.

Four limitations are worth knowing:

• Not every assistant lands in this channel. Google doesn’t publish the full list of recognized sources. GA4 Help names ChatGPT, Gemini, DeepSeek, Copilot and Grok as examples, and the announcement of 13 May 2026 also mentions Claude. Perplexity doesn’t appear in either of these places, and according to some analyses its visits still go to Referral. The list may change, so check the channel description in GA4 Help from time to time.
• Some visits still fall into Referral or Direct. This happens when the assistant doesn’t pass the source or the user copies the address into the browser.
• AI Mode and AI Overviews in Google count as organic traffic. GA4 Help explicitly excludes them from the AI Assistant channel and counts them in Organic Search. What this means for traffic on your site I write about in the guide on Google’s AI Mode and website traffic.
• Visits without a click don’t exist in GA4. If an assistant recommends your company and someone calls without visiting your website, analytics won’t see it.

Step 1: find the channel in the report

The simplest way is to start with a ready-made report:

1. Open GA4 and select Reports, then Acquisition, then Traffic acquisition.
2. In the table, check the Session default channel group column (this is the default dimension of this report).
3. Set the date range from 13 May 2026 to today.
4. Look for the AI Assistant row. If it isn’t there, GA4 did not record such visits in the chosen period, or they went to other channels.

Write down three numbers: Sessions, Engaged sessions and Key events (for example a form submission or a click on a phone number). Menu names may differ slightly depending on the interface version.

Step 2: check what lands in Referral and Direct

The AI Assistant channel is only part of the picture. Now look at the visits GA4 may have classified differently:

1. In the same report, click the Referral row and look at the list of sources.
2. Look for names related to AI assistants (for example the domains of their apps).
3. Remember that Organic Search also contains visits from AI Overviews and AI Mode in Google. GA4 doesn’t separate them from ordinary clicks from search.
4. You can’t break Direct visits down into parts. Just remember that some of them may come from assistants.

ChatGPT appends the parameter utm_source=chatgpt.com to many links in its answers. Thanks to this you will see the source chatgpt.com in reports even when GA4 doesn’t assign the visit to the AI Assistant channel. When the address contains a source without a medium, the session may land in the Unassigned row, so look there too.

Step 3: build your own filter for all assistants

To collect everything in one place, create a simple report in Explorations with the Session source dimension and a filter with the condition “matches regex”. Here is a sample pattern to paste:

.*(chatgpt|openai|perplexity|gemini|copilot|claude|deepseek|grok).*

The .* characters at the beginning and the end are needed because GA4 by default compares the expression against the entire value and is case-sensitive. Thanks to them the filter catches names containing these words, for example chatgpt.com. Adapt it to what you actually see in your data, because source names change. Set the metrics next to it: Sessions, Engaged sessions and Key events. Save the report and open it every month with the same date range.

An example from a small business: a calculation

The example below is made up, only to show the arithmetic. The company builds terraces and has a website with a quote form.

• In the AI Assistant channel and in Referral from assistant domains combined: 40 sessions a month.
• Of those sessions, 3 people submitted the form, so the conversion rate is 3 / 40 = 7.5%.
• The average revenue from an inquiry that ends in a job is determined from the company’s history. Suppose every third inquiry becomes a job worth PLN 12,000.
• From 3 inquiries you get 1 job, that is PLN 12,000 from about 40 visits: on average PLN 300 per AI visit.

It is only an example, but it shows the method. The value of such a visit depends on your industry and on how effective your sales are. With small numbers (a handful of inquiries), one month proves nothing, so look at the trend over 3–6 months. Compare the result with the conversion rate from Google Ads and from organic search. If AI traffic converts similarly or better, it is worth looking after.

What to do when AI traffic is small

Small AI traffic is a normal situation for a local business. Instead of stopping there, you can do three things:

1. Check what the assistants say about your company. Ask them 5 questions your customers would ask and save the answers.
2. Improve what they can cite. I have a separate guide on this: llms.txt and your company’s visibility in ChatGPT. If you want me to review your website for AI bots, I handle this in the AI visibility service.
3. Make sure you are measuring the right events. Without key events (phone, form) you won’t find out whether a visit from AI was worth anything. I describe setting up GA4, Tag Manager and conversions in the Analytics and tags service.

If you don’t know whether your website is suited to being measured and read by AI, start with the free Check your website tool. In 30 seconds it checks, among other things, AI visibility and the technical basics.

Checklist: ChatGPT traffic in GA4

Go through it once a month, in 15 minutes:

1. Open Traffic acquisition and write down the sessions from the AI Assistant channel.
2. Check Referral for the domains of AI assistants.
3. Run your own report with the filter on assistant names and save the result.
4. Note the key events (phone, form) from those visits.
5. Calculate the conversion rate: key events divided by sessions.
6. Compare it with Google Ads and with organic traffic.
7. Ask 3 new customers how they heard about you. Add “AI” to the list of answers.
8. Once a quarter, check in GA4 Help that the list of assistants in the channel hasn’t changed.

Also add one question to your form: “How did you hear about us?”. It is the cheapest way to catch customers whom AI recommended and who called without clicking a link.

Summary

The AI Assistant channel in GA4 makes it easy to see visits from assistants, but it shows only part of them. You will find the rest in Referral, in organic traffic and in your customers’ answers. The most important thing is to look at the same numbers every month and to calculate the conversion rate, not just the number of visits.

If you want to be sure GA4 measures phone calls, forms and visit sources, write to me. I will set up analytics and conversions or check how AI assistants see your website. I reply within 24 hours, and new clients get 20% off their first order.