AI assistants send referral traffic that arrives under their own domains, so it can be separated in GA4 by building a custom channel group matching those referrers. The setup takes about twenty minutes. The important caveat is that a meaningful share of AI-driven visits never carry a referrer at all and land in direct, which means whatever you measure is a floor rather than a total.
Why it is worth separating
AI referral volume is small relative to organic search for most sites, which is exactly why it disappears when left in an aggregate. Two reasons to pull it out anyway.
It converts differently. Published analyses consistently find AI-referred visitors converting well above traditional organic. Averaged into a general figure, that signal vanishes.
The trend matters more than the level. The useful information is direction over quarters, and you cannot see a trend in a segment you never created.
The referrers to capture
The major assistants send traffic under identifiable domains. The practical approach is to match on the recognisable portion rather than maintaining an exact list, since these change:
- ChatGPT and OpenAI properties
- Perplexity
- Claude
- Copilot and Bing's assistant surfaces
- Gemini and Google's AI properties
Use a "contains" match on the distinctive part of each domain rather than an exact match on a full hostname. Subdomains and regional variants appear and change, and an exact-match list quietly stops working without telling you.
Setting it up
- Admin, then Data display, then Channel groups. Create a new group rather than editing the default, which cannot be modified and which you will want intact for comparison.
- Add a channel named something like AI Assistants, positioned above Referral and Organic Search in the ordering. Channel groups evaluate in order, so a rule placed below Referral never fires.
- Define the condition as source matching any of the domains above.
- Save, then wait. Custom channel groups apply going forward and to a limited window of historical data, so do not expect a full history immediately.
- Build a comparison in your reports so you can view AI-referred sessions against organic on conversion rate, engagement, and pages per session.
An alternative that avoids channel group ordering entirely is a simple exploration filtered on session source, which is quicker to produce and easier to share, though less durable.
The limitations, stated honestly
Much of it lands in direct
Assistants running in applications rather than browsers, and links opened in ways that strip the referrer, produce visits with no source attached. Those become direct traffic and are indistinguishable from someone typing your address.
So the figure you build is a floor. Treat it as directional evidence, not a count.
A citation without a click is invisible
Being quoted in an answer that satisfies the user produces no session at all. The value of appearing in an answer is therefore systematically under-measured by any analytics tool, and always will be.
Volumes are small enough to be noisy
Conversion rates calculated on small samples move alarmingly on very few events. Look at quarters, not weeks, and be sceptical of dramatic percentage changes on a small base.
What to do with the data
- Compare conversion rate against organic. If AI-referred visitors convert better, that is an argument for allowing the crawlers that produce them, which connects to local SEO.
- Look at landing pages. Which content is being cited tells you what kind of material earns citations on your site, which is more useful than any general advice.
- Track direction quarterly. The level is less informative than whether it is rising.
- Resist over-reading it. This is a small, partially observable channel. It deserves a segment, not a strategy of its own.
Frequently asked questions
How do I track AI traffic in GA4?
Create a custom channel group under Admin, Data display, Channel groups, with a channel matching the referrer domains of the major AI assistants. Position it above Referral and Organic Search in the ordering, since channel groups evaluate rules in sequence and a rule placed lower will never fire.
Why does some AI traffic show as direct?
Because assistants running inside applications rather than browsers, and links opened in ways that strip the referrer, produce visits with no source information. Those sessions land in direct and cannot be distinguished from someone typing the address, so any AI figure is a floor rather than a total.
Should I use exact match or contains for AI referrer domains?
Contains, matching the distinctive part of each domain. Subdomains and regional variants change over time, and an exact-match list will silently stop capturing traffic without any indication that it has broken.
Will a custom channel group show historical data?
Only partially. Custom channel groups apply going forward and to a limited window of past data, so a full history will not appear immediately after setup. This is a reason to create the segment before you need it rather than when you do.
Is AI referral traffic worth measuring if the volume is small?
Yes, because it typically converts well above traditional organic and because the trend over quarters is more informative than the current level. Left inside an aggregate, both the conversion difference and the direction of travel become invisible.
Can I measure AI citations that do not produce clicks?
Not through analytics. A citation that satisfies the user generates no session, so the value of appearing in an answer is systematically under-measured. Periodically searching your key terms and recording what appears remains the practical approach.
West Coast Media Solutions Inc. builds websites and runs a network of content properties. AI search is changing quickly and much of the published advice in this area is unverified; where a claim is contested we have said so rather than presenting it as settled.
Want a straight read on your AI search visibility?
We will show you what the evidence supports, where your competitors are already appearing, and which work actually pays.
Start a Conversation