For two years, traffic from ChatGPT, Gemini and Claude has been hiding in your GA4 reports, lumped into Referral or Direct. Google just fixed that. GA4 now has a dedicated AI Assistant channel in the Default Channel Group, so you can finally see how much traffic AI tools send you. Here is how to find it, check it is accurate, and actually use it.
Key takeaways
- GA4 added an AI Assistant channel that isolates ChatGPT, Gemini and Claude referrals.
- It appears automatically in acquisition reports, no setup needed.
- Use it to measure AI-referred traffic, not to chase rankings. It is a measurement signal, not an SEO score.
What the AI Assistant channel is
A channel in GA4 is just a bucket of traffic grouped by where it came from: Organic Search, Paid Search, Referral, Direct, and so on. Until now, a click from a ChatGPT answer landed in Referral if the referrer was passed, or in Direct if it was not. Either way you could not isolate it.
The AI Assistant channel groups those visits together. It is part of the Default Channel Group, so it shows up across your standard acquisition reports without any configuration. For the first time, you can answer a simple question your CEO is already asking: how much traffic do we get from AI?
How to find it in GA4
Open your GA4 property and go to Reports, then Acquisition, then Traffic acquisition. In the table, the primary dimension is Session default channel group. You should now see AI Assistant listed alongside Organic Search and the rest. Set your date range to the last 90 days to get a readable trend.
If you want detail, add a secondary dimension of Session source or Session referrer. That breaks the channel down by the specific assistant, so you can see ChatGPT versus Gemini versus Perplexity versus Claude.
Heads up. The channel rolls out per property and may take a few days to appear. If you do not see it yet, it is not broken, it is just not enabled on your property yet.
How to sanity-check the data
Do not trust a brand-new channel blindly. Run two quick checks before you report a number to anyone.
First, look at the trend day by day, not just the 90-day total. A channel that reads zero until a recent date and then jumps is normal for a feature that just launched. You want to report the stable period, not the ramp-up.
Second, cross-check the landing pages. Sessions from AI assistants should land on real, content-rich pages, your guides and comparison articles, not your checkout or a random parameter URL. If the landing pages look wrong, you are probably looking at misclassified traffic, not genuine AI referrals.
Treat the AI Assistant channel as a new measurement signal to validate, not a number to forward to the board on day one. MyDigipal
What the numbers tell you, and what they do not
This channel tells you volume and behaviour: how many sessions arrive from AI, which pages they land on, and how they engage. That is genuinely useful. It shows whether your content is being cited and clicked inside AI answers, and which topics pull people through.
What it does not tell you is your visibility inside the AI answer itself, the times you were cited but not clicked. Click-through from AI assistants is low by nature, because the answer often satisfies the user on the spot. So read this channel as a floor on your AI presence, not the full picture. Pair it with your own prompt testing to see where you actually show up.
Turn it into action
Once the data is clean, do three things. Mark the pages AI assistants send traffic to, and double down on that format, those are the ones being cited. Add the AI Assistant channel to your monthly reporting so the trend is tracked, not anecdotal. And feed the winning topics back into your content plan, so you write more of what AI tools already pull people in on.
Tracking and reporting
Want AI traffic measured properly, not guessed?
We set up clean GA4 tracking and reports your team can act on, including AI-referred traffic, so you know what is working across search, AI and paid in one place.
The AI Assistant channel is a small change with a big upside: you can finally put a number on a question everyone has been guessing at. Find it, validate it, track it monthly, and use it to write more of what AI already rewards. If you want help wiring it into reporting that drives decisions, our SEO and LLMO team can help.
