AI referral traffic
AI referral traffic is the sessions arriving on your site from links inside AI assistant answers — chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar. It is the one AI-visibility signal that lands in your own analytics, which makes it the easiest to report and the easiest to misread, because it counts only the answers someone clicked out of.
How to isolate it
Referrer hostnames are the practical handle. Build a channel group or a segment matching the assistant domains you care about — chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai — and keep it in one place so the definition does not drift between reports.
Two caveats worth writing into the report footnote. Referrers get stripped or rewritten in some clients, so the segment undercounts. And Google's AI Overviews and AI Mode sit inside Search, so their clicks arrive as ordinary organic google.com traffic and cannot be separated in standard analytics at all. Any figure you quote for "AI traffic" therefore excludes the surface with the largest reach.
What the numbers usually look like
Small in absolute terms, different in character. Two patterns show up consistently enough to plan around:
Sessions are a fraction of organic. For most catalog sites AI referrals are a low single-digit percentage of total sessions. Anyone presenting it as a channel that will replace organic search this year is ahead of the data.
The visits behave unusually. Sessions arriving from an assistant tend to land deeper in the site and arrive later in the buying process, because the assistant already did the orientation work. In B2B that shows up as more spec-page and part-number landings and fewer category-page browses. Judge the channel on what those sessions do, not on their volume.
The undercount problem
Referral traffic measures clicks. AI visibility is about answers, and most answers do not produce a click. A buyer who asks an assistant which coupling fits a 1-7/16 shaft, gets your part number in the answer, and then types that part number into your site search shows up in your analytics as direct or as internal search. Nothing attributes it to the AI surface.
So treat referral traffic as a lower bound and pair it with the instruments that see the other side: retrieval-crawler hits in server logs, and a fixed prompt testing panel. Also worth watching is the ratio of branded to unbranded queries in Search Console — the assistant-mediated path often converts an unbranded need into a branded search later.
Before you optimize for it
The instinct on seeing a small AI referral number is to build content aimed at assistants. For a distributor that is usually the wrong first move, because the constraint is upstream: pages that lack the attribute a buyer asked about cannot be cited no matter how the content is packaged.
A reasonable sequence is to check crawler access first (are OAI-SearchBot, PerplexityBot and ClaudeBot actually allowed and getting HTML), then coverage on the attributes buyers filter on, then packaging. Reporting comes last, and should be framed as a trend line rather than a monthly number.
Frequently asked questions
How do I track AI referral traffic in GA4?
Create a custom channel group or segment that matches assistant referrer hostnames such as chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. Keep the definition documented, since it will need updating as new assistants appear and as referrer behaviour changes.
Why is AI referral traffic so low compared to the hype?
Because it only counts clicks. Most AI answers resolve the question in the answer itself, and Google's AI Overviews and AI Mode clicks arrive as ordinary organic google.com traffic that analytics cannot separate. The metric is a floor, not a measure.
Does AI referral traffic convert better than organic?
It often behaves differently — deeper landing pages, later-stage intent — but published conversion multiples floating around the industry come from small unaudited samples and should not be planned against. Measure your own, segment by landing page type, and be honest about sample size.
Can we see AI Overviews traffic separately?
Not in standard analytics. AI Overviews and AI Mode are surfaces within Google Search, so their clicks are attributed to organic google.com. Search Console reporting has evolved on this, but there is no clean historical split, which is why crawler logs and prompt panels remain necessary.