How to measure whether AI assistants mention your store
Last reviewed
Short answer
There is no rank in an AI answer to track, so measurement means sampling. Build a fixed set of buyer-intent prompts for your category, run them on a schedule across assistants, and record how often your store is mentioned. The output is a mention rate over time, not a position.
Why there is no rank to track
Classic rank tracking works because a query returns an ordered list that is broadly stable across users. An AI answer is generated, varies between runs, and often names two or three products rather than ten. There is no position five to hold.
So anyone selling you an AI rank is selling a number they invented. The measurable thing is frequency: across a fixed set of questions, how often do you get named at all. That is a real signal and it moves.
Building a prompt set
Write the questions a buyer would actually type, not the keywords you rank for. "Best dishwasher safe coffee mug for the office" rather than "ceramic mug". Thirty to fifty prompts covering your main categories, price points and use cases is enough to see movement without the cost getting silly.
Freeze the set. The single biggest mistake in this kind of measurement is changing the questions between runs, which makes every comparison meaningless. Add new prompts as a separate cohort if you must.
Reading the result honestly
Report it as a sampled rate with the sample size attached: six mentions out of fifty prompts this week, up from three. Never as a percentage without the denominator, and never as a rank. Expect noise, especially at small sample sizes, and look at the trend across several runs rather than reacting to one.
Pair it with something you control. If you fixed 40 product titles in week one, you want the mention rate line and the readiness score line on the same chart. That is the closest thing to attribution available here, and it is still correlational.
What good tooling costs
Dedicated AI visibility platforms price in the hundreds of dollars a month. The underlying API cost of running fifty prompts weekly across three assistants is a few dollars. You are paying for the dashboard and the roll-up, which is a fair thing to sell but worth knowing.
Entitled includes weekly sampling across ChatGPT, Perplexity and Claude, framed as directional rather than as a rank. See the sampling view in the demo.
Frequently asked questions
Can I track my rank in ChatGPT?
No, because there is no rank. AI answers are generated and vary between runs. The honest metric is a mention rate across a fixed prompt set, sampled on a schedule and reported with its sample size.
How many prompts do I need?
Thirty to fifty covering your main categories and use cases is a reasonable starting point. Fewer than about twenty and normal variation will swamp any real change.
How often should I run it?
Weekly is a sensible default. Daily adds cost and noise without adding signal, since the underlying data changes slowly. What matters more than frequency is keeping the prompt set fixed.
Sources
Keep reading
Score your own catalog
Entitled grades every product against the rules on this page, explains each failure, and fixes them with AI you approve. The score is free forever.