Use GA4 to see which AI engines send visitors. Build a filtered view of referral traffic from known AI hostnames — chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar — then group them into a channel. No paid tool is required; the referrer patterns tell you the source.
You can do this in GA4
You can see which AI engines send you traffic using GA4 alone, no paid tool needed. Most AI surfaces pass a referrer when a user clicks through to your site, and that referrer names the engine. By filtering for those hostnames, you can separate ChatGPT visitors from Perplexity visitors from ordinary search.
It is not perfect, because some engines strip or obscure the referrer, but it captures enough to show you which surfaces are actually working and which are not yet sending anyone.
Know the referrer patterns
Each engine tends to send a recognizable hostname. Learn the common ones and you can spot them in your reports.
- ChatGPT: chatgpt.com
- Perplexity: perplexity.ai
- Gemini: gemini.google.com
- Copilot: copilot.microsoft.com and Bing chat paths
- Claude: claude.ai
This is the same source list used more broadly in how to track AI referral traffic. Here the aim is narrower: attributing each visit to a specific engine so you know who your best AI referrer is.
Build the view in GA4
Set it up once and reuse it. In GA4, explore your traffic acquisition report, then filter session source or referrer by the hostnames above. Save that as a comparison or a custom exploration so you do not rebuild it each time.
For a cleaner ongoing read, create a custom channel group that bundles those AI hostnames into a single "AI" channel, with sub-rows per engine. Then AI traffic shows up beside Organic and Direct in your standard reports.
Handle the gaps honestly
Some AI traffic will slip through. Certain engines and app contexts pass no referrer, so those visits land in Direct and cannot be attributed. Do not pretend otherwise. Report the AI traffic you can identify as a floor, not a full count.
You can narrow the gap a little with UTM tags on links you place inside your own content or profiles, but you cannot tag links an engine generates. Accept a known undercount rather than inventing precision you do not have.
Connect traffic back to mentions
Referral traffic is only half the story. It tells you which engines send visitors, but not how often you are named in answers that do not get clicked. Pair this with measuring AI share of voice so you see both presence and resulting clicks.
An engine can cite you constantly and send few visitors, especially when the answer fully satisfies the user. Reading traffic and mentions together keeps you from misjudging an engine that is building your visibility without yet sending clicks.
Make it a standing metric
Once the view exists, check it on the same cadence as your other numbers and watch the per-engine trend. Rising traffic from a given engine tells you where your content is landing and where to concentrate next.
Fold the result into your regular scorecard alongside the GEO KPIs that actually matter. Which engines send traffic, and whether that traffic is growing, is one of the few AI visibility numbers that shows up cleanly in your own analytics.
Reading the per-engine picture
Once you can attribute traffic by engine, the interesting work is interpreting the mix. Different engines send different kinds of visitors, and knowing which one drives your results tells you where to concentrate.
Watch a few things over time. Which engine sends the most visitors, and is that share growing or shrinking. Which engine sends visitors who actually do something once they land, since raw sessions matter less than what follows. And whether a new engine starts appearing, which signals your content is being picked up on a surface you were not watching.
The patterns guide your effort. If Perplexity sends engaged visitors and Copilot sends almost none, that tells you where your content is landing and where a gap sits. If one engine suddenly climbs, look at what you published or earned just before, because something you did is resonating there. Reading the per-engine picture turns a flat traffic number into a map of where your visibility is strongest, so you can double down on the surfaces already working and shore up the ones that are not.
Questions people ask
No. Many do, which is how GA4 can attribute visits to ChatGPT, Perplexity, Gemini, and others by hostname. But some engines and in-app contexts strip or omit the referrer, so those visits land in Direct and cannot be traced. Treat your identified AI traffic as a floor rather than a complete count, and report it honestly as an undercount.
No. GA4 is enough. Filter your traffic acquisition report by the known AI hostnames, or build a custom channel group that bundles them into an AI channel with a sub-row per engine. Paid tools can add convenience and catch some edge cases, but the core method is just recognizing referrer patterns in analytics you already have.
Because the generated answer often satisfies the user without a click. Being named still builds awareness and trust even when no visit follows. That is why you should read referral traffic alongside share of voice: traffic shows who clicks through, while share of voice shows how often you are mentioned. An engine can grow your visibility long before it sends meaningful traffic.
Want this done for you?
Want to know if AI is recommending you or your competitor? Get an AI visibility audit from Acromatico.
Get a free AI Visibility Audit →