Learn
How to find AI agent traffic in Google Analytics
Google Analytics can show you people who arrived from an assistant, because those are real browser sessions carrying a referrer. It cannot show you agent requests themselves, because most never run the JavaScript that reports a pageview. Analytics answers the referral question and your server logs answer the agent question, and confusing the two is how stores conclude that nothing is happening.
Google Analytics can show you the people who arrived from an assistant. It cannot show you the agents themselves. Those are two different questions, and treating the first as an answer to the second is how a store concludes that nothing is happening.
Why the agent itself is invisible there
Analytics is measured by a script that runs in the visitor’s browser. An agent that fetches your markup and reads it never executes that script, so the request is real, served by your infrastructure, and entirely absent from your reports.
This is not a configuration problem and there is no setting that fixes it. It is a property of where the measurement happens. The requests exist in your server logs, which is why server logs are the honest source for this question.
What you can see: the referred human
When an assistant sends a shopper to your checkout, that is a genuine browser session and it does appear. Build a segment for it rather than looking for a built in channel, because assistant referrals are commonly bucketed as direct or as generic referral traffic.
- Create an exploration segmented on session source, matching the assistant hostnames you care about. Keep it as a pattern rather than a fixed list, since the set changes.
- Add a second segment for sessions with no referrer on a mobile viewport arriving straight onto a product or cart URL. This is a weak signal on its own and a useful corroborator alongside the first.
- Compare conversion and revenue per session against your baseline rather than reading the raw count. The count is understated for reasons you cannot fix. The ratio is still informative.
Why the number will be too low
Referrers are lost easily. A redirect chain drops them, a consent wall can drop them, a strict referrer policy strips them, and a purchase completing in an in-app browser may not present one at all. Every one of those failures pushes the number down, never up.
So treat what you measure as a floor. If you need to know whether the number is worth acting on, how much of your traffic comes from AI agents covers what the figure can and cannot support.
Do this before you need it
The segment costs an afternoon and cannot be backfilled. The moment somebody asks how large this channel already was, the answer is limited to whatever you happened to be capturing at the time, and everything before that is unrecoverable.
Questions
Why does agent traffic not show up as a channel?
Because analytics is measured by a script that runs in a browser. An agent fetching your markup never runs it, so the visit is real and unrecorded. What you can see is the human who arrives afterwards, which is a different and smaller thing.
Will referral data tell me which assistant sent someone?
Sometimes, and it is fragile. Referrers get stripped by redirect chains, consent walls and referrer policies, and an in-app browser may not pass one at all. Treat a present referrer as a lower bound rather than a count.
Should I build this reporting now or wait for the volume?
Build the segment now, because you cannot backfill it. Attribution you did not capture is gone, and the moment you want this data is the moment somebody asks how big it already was.