We count what we can, model what we can't, and mark what no one can see yet.
What's at stake for you
Why this lands on your desk
You are the skeptic in the room, and you should be. Everyone claims they can track AI traffic; almost no one can, and you are the one who has to defend the numbers.
You do not need magic. You need to know exactly how each figure was derived, and where the limits are, before it goes in a report with your name on it.
What you get
Numbers you can put your name on.
A measurement frame: every number labeled measured, estimated, or unknown
Citations counted directly from live engine answers
Modelled figures (crawls, AI-referred visits) marked as modelled
Trends over time — AI share of voice, branded vs unbranded, referral movement
How it fits your process
Where we stop, and what stays yours
Real versus modelled data is labeled explicitly; coverage numbers are never extrapolated.
Every metric carries how it was derived, so you can defend it rather than take it on faith.
It's never a PDF once a week — it's a live dashboard you can audit any time.
The free GEO audit
See exactly where AI engines are ignoring you.
Seven gaps, measured on your own domain — before you pay anything.
FAQ
Questions from your seat
Can you even track AI traffic?
Not perfectly — no one can, and the engines deliberately hide their agents. We deploy the best measurement available and split every number into measured (counted from live answers), estimated (modelled until logs connect), and unknown (what the engines do not expose).
What's actually measured versus estimated?
Citations are measured — counted from real engine answers. Crawls and AI-referred visits are estimated — modelled proportionally until server logs are connected. We never present one as the other.