AI visibility tracking, measured the same way every time.

AI visibility tracking runs the same fixed set of customer prompts through the same AI assistants on a schedule and records each time whether your business is named, so that a real change can be told apart from noise. It’s the AI counterpart of a rank grid, with one important difference: the thing being measured doesn’t hold still.

Why a single check misleads

An AI assistant writes each answer fresh, so the same question can return different companies on different days. Ask which electrician in Carlsbad can upgrade a panel for an EV charger and you might be named this morning and absent this afternoon. A screenshot of one good answer proves almost nothing, and so does a screenshot of one bad one. Anyone showing you a single result as evidence, in either direction, is showing you one sample of something that varies.

Why the prompt set stays fixed

Small changes in wording change the answer. “Electrician in Carlsbad” and “electrician near Carlsbad for a panel upgrade” are two different tests, so the prompts stay word-for-word the same. We agree on the set at the start, and it usually comes out of the AI visibility audit. If you add a service or a city later, I add the new prompts as new lines and leave the old ones untouched, because editing them would make the history stop being comparable.

What gets counted

I run each prompt several times per assistant and write the result as a count, in the form “named in three runs out of ten”. That makes the measure a rate, and a rate can be compared from one period to the next, which a single yes or no can’t.

Alongside it I record which competitors were named, which sources were cited, and whether the facts given about your company were right.

When the prompts are re-run

On a regular schedule, and again about two weeks after any change I make. That’s the same rhythm I use for rank grids, and for the same reason: one change at a time, measured before and after. Cause is harder to pin down here than with rankings. The companies behind these assistants update them without notice, and a shift in your numbers may have nothing to do with anything I did, so a movement I can’t tie to a change goes into the report marked that way.

What tracking can’t tell you

The numbers describe how often you’re named on the prompts tested, by the assistants in the set, and nothing beyond that. A customer can ask a question that isn’t on the list, and a rate that is rising can’t be guaranteed to keep rising. The counts also say nothing about how those assistants differ from one another, which belongs to GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization).

Reported next to rankings and calls

I put the AI figures in the same report as your map rankings and your phone calls, not in a separate document with its own vocabulary. Calls are the number that pays your crew, so the AI figures sit beside call tracking where you can see whether they move together. I don’t credit a call to an AI answer unless there’s evidence for it.