Menu

How to track AI visibility for a local business

By the Pickgap team · Updated Oct 15, 2026 · 4 min read

Track AI visibility by asking the questions your customers ask, in your city, on each AI surface, several times each, and recording how often you're named. That share is your pick rate. Because answers change from one ask to the next, one search proves nothing; you need enough answers to see a pattern, a baseline before you make changes, and the same test repeated afterwards.

Why AI visibility is hard to measure

A Google rank is fairly stable: search the same thing from the same place and you see roughly the same order. AI answers aren't like that. Ask ChatGPT "best plumber in Mesa" three times and you may get three different sets of names. The answer also depends on the exact wording, the searcher's location, and which AI you ask.

So the question isn't "where do I rank?" It's "how often am I named, across the questions customers ask?"

The pick rate

The simplest useful measure is the pick rate: the share of AI answers that name your business.

If you ask 25 questions twice each on 5 AI surfaces, you get 250 answers. If 30 of them name you, your pick rate is 12%.

Report it with the sample size ("30 of 250 answers") and, ideally, a noise range, because a pick rate from 20 answers is much less certain than one from 250. See the glossary entry on pick rate.

Compare it with the business that's named most in your category and city. The distance between you is your gap, and closing it is the goal.

Step 1: Choose the questions

Use questions written the way customers ask AI, not keyword fragments:

A mix matters: businesses are often named for some types of question and missing from others.

Step 2: Choose the AI surfaces

The main surfaces for local recommendations are ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode. They use different sources and pick differently, so measure each one separately.

Step 3: Repeat each question

Ask each question more than once on each surface. Two runs is a reasonable minimum to smooth out the random variation; more is better if you can.

Step 4: Record everything

For each answer, record:

The competitors and sources are as valuable as your own result: they show who you're up against and where AI looks.

Step 5: Re-test on a schedule

Treat a change as real only when it's bigger than the noise. Moving from 12% to 14% on 50 answers is noise; moving from 12% to 28% on 250 is a change.

Doing it by hand vs. with a tool

You can run a small version by hand; see how to check if ChatGPT recommends your business. It works for a quick look, but a full test of 25 questions, 2 runs each, on 5 surfaces is 250 answers, which takes hours by hand and is hard to keep consistent.

Pickgap runs that test automatically as a searcher in your city, records every answer and source, and reports your pick rate with its noise range, the businesses picked instead, and a fix list. A weekly pulse re-asks your top 5 questions. Plans start at $29 a month; the free check runs 5 questions on all 5 surfaces with no signup. The full method is public at How we measure.

Also track the traffic

Pick rate tells you whether you're named. Your analytics tell you whether that sends visits:

See how do I know if AI sends me customers?

Questions

How often should I check my AI visibility?
Run a full test before making changes and every 4 to 8 weeks after. A short weekly check of your most important questions catches big swings in between.
Why do I see different results than my report?
AI answers vary by wording, location, account settings and plain chance. A single search on your phone is one sample. A standardized test of many answers shows the pattern, not what every person sees.