
Business owners keep hearing that AI assistants send visitors to websites. Google Analytics 4 can show whether yours is getting any, and jamn checks that report on its own site as well.
Short answer: In Google Analytics 4, visits from AI assistants such as ChatGPT can appear in the Traffic acquisition report under the AI Assistant channel. The source shows as the tool’s domain, for example chatgpt.com, and the medium shows as ai-assistant. The counts are often small, and some AI visits arrive with no referrer and get counted as Direct.
Where AI referrals appear in GA4
GA4 has a native AI Assistant channel, and it needs no manual setup. Here is how to find it:
- Open the Traffic acquisition report under Acquisition in the Reports menu.
- Look for the AI Assistant row in the default table, which groups sessions by Session default channel group.
- Switch the dimension dropdown to Session source / medium to see which tool sent each visit.
A real entry looks like chatgpt.com / ai-assistant.
What the report looks like with real numbers
jamn’s own property shows 654 sessions for July 7 to October 4, 2026. These sample rows come from that report, so you can see how an AI referral sits beside other sources:
| Session source / medium | Sessions |
| google / organic | 47 |
| ig / social | 35 |
| chatgpt.com / ai-assistant | 1 |
The single chatgpt.com session was an engaged one with 12 seconds of average engagement time, and it makes up 0.15% of all sessions. One session says nothing about a trend. The row shows what a recorded ChatGPT referral looks like in this report. It doesn’t prove that every relevant AI assistant is being tracked or that the reporting is complete.
What the native channel may miss
Two gaps are worth knowing about. The first is coverage. Google’s documentation discusses Perplexity when it explains how to build custom AI-assistant channel groups, but that doesn’t settle whether Perplexity is missing from the native channel’s list of recognized sources. Check your own Session source / medium report first. If relevant AI sources show up as ordinary referrals, or aren’t grouped the way you want, a custom channel group (covered below) lets you group them your own way in your reports.
The second gap is Direct. Some visits from AI apps arrive with no referrer and land there. For example, a link copied from a chat app and pasted into a browser has no referrer, so GA4 records the visit as (direct) / (none). Direct generally means GA4 has no usable source information for the session.
Internal and automated traffic is a separate issue. Unfiltered internal traffic can contaminate your reports, and automated traffic can distort them, but neither necessarily lands in Direct. Set up an internal traffic filter in GA4 as general cleanup for the whole property, and treat it as its own task.
How to read very small numbers
Counts in the single digits can’t support conclusions about growth or decline. Two things are worth watching instead: whether the same source returns week after week, and which landing pages it reaches. A source that returns in two different months and lands on the same post both times is worth investigating. Repeat referrals are a reason to look closer, and they don’t prove that an AI system consistently recommends that page. A single visit to the homepage tells you much less.
Tag the links you control
UTM parameters on links you place yourself, such as in an email newsletter, keep those visits separate from AI referrals. A newsletter link might look like this:
https://example.com/blog/post/?utm_source=newsletter&utm_medium=email&utm_campaign=fall-update
jamn uses lowercase values with hyphens in every UTM so one campaign never appears under two spellings. You can’t add UTM parameters to links that an external AI service generates on its own. The Traffic acquisition report is a useful starting point for seeing those visits, and landing page reports and Explorations let you dig further into the same sessions.
Optional: a custom channel group for tools GA4 doesn’t group
If a tool you care about shows up as a plain referral, a custom channel group can catch it in your reports. A custom group is an alternative grouping that sits beside GA4’s default one. It doesn’t change the default channel group or the native AI Assistant channel. In Admin, open Channel groups under Data display and create a new group. Then add a channel with a Source condition that matches a regular expression. A simple pattern for Perplexity:
.*perplexity.*
To cover a second tool, separate the names with a pipe character, for example .*perplexity.*|.*other-tool.*. Place the new channel above Referral in the channel order, since GA4 assigns each session to the first channel that matches. GA4 menu names change often, so match these steps to the labels in your own property.
What to do with the results
When a recurring AI source lands on a specific post, update that post and put a direct answer near the top. How AI summaries choose their sources is covered in jamn’s post on query fan-out.
For the Google side of AI visibility, Search Console’s generative AI performance report shows how often your pages appear in Google’s generative AI features. That covers AI Overviews and AI Mode in Search, and generative AI features in Discover. Google launched the report on June 3, 2026, and it reports impressions only, so it can’t tell you who clicked. The same data also stays in the overall Performance report. Google’s documentation says the worldwide rollout finished on August 31, 2026, though a property with too few impressions in these features may still see no report. The report measures visibility in Google’s own generative AI features. GA4’s AI Assistant channel covers a different set of visits, the ones that come from tools like ChatGPT.


