The situation
Leadership keeps asking how much traffic comes from ChatGPT, Perplexity, and AI Overviews, and the honest answer in the room right now is "we think it's growing." That is not a number. It is not even a trend - it is a feeling dressed up as an update.
Why it's a real problem
A channel nobody measures is a channel nobody can defend budget for, explain a spike or drop in, or compare against paid and organic. Meanwhile the questions your buyers used to type into a search bar are increasingly being asked inside a chat window you have no visibility into at all.
What we implement
We segment AI referrers in your existing analytics, analyze AI crawler activity in your server logs, and add labeled dark-traffic inference only where direct measurement genuinely runs out - together, AI traffic measurement, built on top of the stack you already run rather than a separate tool to babysit.
What you get
- A real, referrer-based floor number for AI-driven sessions, refreshed on your existing reporting cadence - not a special one-time pull nobody repeats.
- A clear line between what is measured and what is inferred, so the number survives the first follow-up question instead of falling apart.
- Crawler-log evidence of retrieval demand, which typically shows up before referred sessions do - an early signal, not an afterthought.
Illustrative scenario
A B2B software company that assumed AI traffic was negligible might find, once referrers and crawler logs are segmented, that retrieval demand from AI bots was already running weeks ahead of any referred sessions showing up in analytics. (Illustrative scenario - not a measured result.) What your own numbers show is the point of doing this, not this example.
The honest limit, stated up front: engines do not always pass a referrer, and app-embedded experiences can strip it entirely. Every report we hand over says which number is measured and which is inference - that discipline is what lets it be quoted upward.