How to Measure AI Visibility So the Number Actually Means Something

An AI-visibility chart made the rounds on LinkedIn last week, one clean line climbing from 250 mentions a day to 1,700 in two weeks. The person who posted it was thrilled, and at a glance anyone would be. The comments were less impressed, because one of them asked how many queries the tool had run to reach that number.

That question is the whole game. A mention count goes up when a business gets recommended more often, and it also climbs when the tool simply asks more questions. Widen the query pool and the line rises on its own, with the business sitting exactly where it started. Most tools never show that pool, so real growth and a bigger sample look identical from the outside.

It helps to name the thing being measured. AI visibility is how often an assistant like ChatGPT or Google AI points a customer toward a specific business when it answers a question. Marketers arriving from search are used to rankings that sit still and can be audited. AI moved the surface and the scoreboard at the same time, which is part of what separates GEO from old-school SEO. It also created a measurement problem, because the easiest number to report is the easiest one to inflate.

Why an AI visibility number climbs on its own

Nobody has to lie for this to happen. A tool chooses which questions to ask, and more questions mean more chances to appear, so the count grows even when the business has done nothing new. Run the same prompts a hundred times a day, then push it to seven hundred, and the chart shows momentum that lived entirely in the sampling.

This is why a raw AI mention count sits close to a vanity metric. It answers “how hard did the tool look” far better than “how often does an assistant recommend this business.” I am still learning how each platform makes its picks, and I will say that plainly. One thing is already clear. A number that jumps when the tool works harder is describing the tool, and it tells almost nothing about the market.

A number worth trusting reads the same at any scale

A figure worth trusting behaves differently from a running total. It holds roughly steady no matter how many questions get thrown at it. Ask a hundred, ask a thousand, and an honest AI visibility metric lands in about the same place, because it reflects a real position and the sample size barely touches it.

That is the test I keep returning to. When a bigger sample changes the headline, the headline was never about the business in the first place.

Share of voice against real competitors

At Reachd, the figure I care about is share. For any business, we look at how often it appears in AI answers next to the competitors it actually fights for customers with, across a broad set of real questions. It works like share of voice inside the answer itself, measured on the same ground for everyone.

Share survives the volume trick by design. Adding more questions lifts the business’s count and every rival’s count together, so the proportion barely moves. Padding the sample stops helping, because it inflates both sides of the ratio at once. That is why share reads the same at a hundred queries or a thousand, and it is the one figure a tool can’t pad from the shadows.

The set of questions matters as much as the math. Real customers never ask the same way twice, so the pool keeps shifting to match how people actually talk to ChatGPT and Google AI. To follow share over time, we line up like with like, comparing the same kinds of questions from one month to the next. A real gain in position then shows up as a real gain, and a fresh batch of easier or harder prompts can’t pass itself off as progress.

One question that tells you if a visibility chart is honest

Spotting a number that can’t be trusted is faster than proving one that can. Actually re-measuring visibility takes a real run of queries over time, and that sustained work is what a tool is for. Judging whether someone else’s chart deserves trust takes a single question. Is this an absolute count or a share against real competitors, and do the queries behind it stay the same from one month to the next. A climbing absolute number with a hidden query pool explains itself, and a share measured against the same rivals is the kind that holds up.

An honest tool should be able to show the questions behind its numbers, and it should lean on share rather than a raw count. That is how we built AI brand monitoring at Reachd, tracking a business’s share against its real competitors over time instead of a mention tally that grows with the sample. For a snapshot of where a business stands today, a free AI visibility check shows who the assistants name in its category right now.

A rising mention count feels like progress, and that is exactly why it sells. The number that earns attention is the one that stays honest when the sample doubles. When a business is genuinely recommended more often than the rivals showing up in its place, that shows up as share, and share is the part of the story worth telling.