How do you measure visibility in AI search?
Short answer
Run a fixed set of real buyer prompts across several assistants, repeat each one, and report how often you are cited as a share of answers. Track AI referral traffic in analytics alongside it, knowing it undercounts. A single screenshot of one answer is not a measurement.
Why one test means nothing
Assistants generate a different answer each time. One result is a draw from a distribution. Measuring visibility means sampling repeatedly and reporting frequency, the same way a poll reports a share rather than one interview.
What to track
- A fixed panel of 15 to 30 buyer prompts, chosen before any work starts
- Citation frequency per assistant and per prompt
- Which sources the assistants used instead of you
- AI referral sessions in analytics, as a floor rather than a total
What cannot be measured
Revenue caused by an assistant's citation cannot currently be attributed. Anyone reporting it as a precise figure is estimating, and should say so.
Send the problem. Get a scoped proposal.
Within 24 hours you get a fixed scope, a delivery window and a price, or a short note saying we are not the right agency for it.