How to measure AI visibility, and why one measurement is not enough
8 min read · Updated: 20 Sep 2026
Contents
Why answers change
Ask the same engine the same question twice and it may suggest different brands. Model updates, web search and location all change the answer.
That is why a single screenshot proves nothing about a brand's visibility.
Pick the right questions
Write questions a customer might really ask: product searches, comparisons, local searches and brand questions. Questions that include the brand name make results look better than they are.
Fix the question set at the start of a period so you can compare week to week.
Decide on engines and repeats
Pick the engines your client's audience uses. Asking each question more than once reduces the effect of lucky results.
Always read the result with its sample: how many questions, engines, repeats and which dates.
Which numbers to look at
Mention rate shows in what share of answers the brand appears, position shows how early it appears. Share of voice compares you with the competitors you track.
Citations show which pages describe your brand; that is often where the fix is.
Follow the trend
Look at a weekly or monthly series instead of a single measurement. Swings of a few points are normal; a lasting drop or rise is what matters.
To see the effect of a fix, measure before and after with the same question set.
Check what this article covers on your own site.
Check your site