Here's the answer: you measure it yourself, monthly, because no platform will email you when it stops recommending you. The instrument is simple. Write the twenty questions your buyers actually ask, put them to the AI engines in fresh sessions every month, and log four things per answer: were you named, where you ranked, whether what it said is true, and which sources it cited. That last column is your work order. It names exactly where the record needs fixing.
The tricky part is the traffic, because most of it hides. Many people read an AI's answer, never click anything, and later type your name straight into a browser. Your analytics call that a direct visit. It was really an AI recommendation wearing a disguise. So you triangulate: track the visible AI referrals as your floor, watch whether searches for your brand name rise alongside your answer visibility, and then use the one honest instrument available at any size: ask buyers "How did you first hear about us?" with an AI assistant listed as an option. Platforms inflate themselves. Customers tell the truth.
Why bother with all this? Because the visitors AI sends convert at multiples of normal search traffic. They arrive late in their decision, already sold by the answer. A hundred AI visits can outweigh a thousand ordinary ones in actual revenue, and if you only count raw clicks, you'll starve your highest-grade channel.
The full scoreboard is Volume 4 of our Digital Stack series. It's also, at market scale, exactly what our Diagnostic does.
What gets measured gets cited.