An AI visibility score is a summary measure of how often an AI assistant names a business across a defined set of questions, on named platforms, over a stated number of runs. Without those three disclosures it is a number rather than a score.
- A score is a summary of a sample. Its entire value rests on what was sampled, where, and how many times.
- The three disclosures are non-negotiable: the question set, the platform list, the runs per question. Missing any one makes the number unreadable.
- Small samples cannot be percentages. Twelve questions produce a fraction. A decimal point on twelve observations is a decorative choice, not a precise one.
- You can get one free. Ours costs nothing, uses a published method, and the findings are yours whether or not we ever speak.
The honest boundaries
It can tell you presence. Across a defined set of questions, how often were you named? That is a real, checkable quantity and it is the only thing most scores are actually measuring.
It can tell you direction over time. Same questions, same platforms, same method, repeated — the movement is meaningful even when the absolute number is rough. This is where the value is.
It cannot tell you a rank. There is no position to hold in a generated answer. Assistants do not maintain an ordered list, and the order they happen to mention names in is not stable between runs. Any product showing you a rank is showing you a coincidence.
It cannot tell you volume. No assistant publishes how many people asked a given question. Where a tool shows “prompt volume,” it is modelling, and the model is not disclosed.
And it cannot be precise on a small sample. Precision comes from observation count. A twelve-question screen tells you roughly where you stand; it does not support claims about a two-point move, and we will not make them.
What it is good for is triage. A score of 1/12 and a score of 8/12 call for completely different work — the first is almost always a structural problem sitting below the content layer, the second is a competitive one. Knowing which situation you are in before anyone quotes you is most of the value, and it is the part a vendor has the least incentive to make clear.
One more limit worth stating. A score measures presence, not consequence. Being named in an answer is not the same as being called, and a rising X/12 alongside a flat phone is a real result that means something specific, usually that the visibility work is landing and the conversion path is where the problem actually lives. A provider who only ever shows you the score is showing you half the picture.
Where this term ends and the next one begins
Four questions before you believe a number
1. What were the questions? Ask to see the whole set. Questions like “best {your company name}” inflate a score to no purpose, a brand query proves only that you exist. The set should be what a buyer asks before they know who you are.
2. Which platforms, in what state? Google AI Overviews, ChatGPT and Perplexity behave differently and should be reported separately. A blended number that hides a zero on one platform is hiding the most useful thing in the report. And a score run from a logged-in account with memory on is measuring the analyst, not the market.
3. How many runs per question? One run is a coin flip. Generated answers vary; a business named once out of three is in a different position from one named three times out of three, and a single-run score cannot tell them apart.
4. What does it refuse to claim? This is the fastest test. A number with no stated limits was not built by someone worried about being wrong. Ours refuses to be a percentage at twelve questions, refuses to report rank, and refuses to narrate small movements as progress.
This page defines a metric. It does not sell anything. The scoring is free and unattached to a pitch: get your Visibility Check here. What it looks like when the work behind the score is run as an engagement is on the AI Visibility service page.
Want the one-paragraph version? The Visibility Glossary keeps short definitions; this page keeps the practice.
Fair questions about AI visibility scores
Why is yours out of 12 and not a percentage?
Because twelve questions cannot support a percentage without implying a precision that is not there. A fraction keeps the sample size visible in the number itself, 7/12 tells you both the result and how much evidence sits behind it, where “58%” hides the second half. Larger samples can carry a percentage, and when we run one we report the confidence interval with it.
Is a low score bad news?
It is usually the most useful thing in the report, because a zero is diagnostic. Most low scores trace back to something structural and fixable, a blocked crawler, contradictory business details, no page answering the question, thin corroboration, rather than to anything mysterious about AI. The score's job is to point at which one.
How often should it be re-run?
Monthly for the twelve-question screen, and read as a trend rather than a reading. Answers shift every few days, so anything more frequent is measuring weather. Compare against the same question set every time; changing the questions resets the baseline and makes the comparison meaningless.
Can I run this myself?
Yes, and we would rather you understood the method than took our word for it. Open a clean, logged-out session with memory off, ask twelve real buying questions three times each across the three assistants, and count where you are named in at least two of three. The published protocol has the details. Doing it once yourself makes you much harder to sell to.
What's the catch with the free check?
There isn't one worth worrying about. It takes us a couple of hours, you get the score, the screenshots and the first recommended fix, and the findings are yours whether you hire anyone or not. We do it because most owners have never seen this data and the ones who have it tend to ask better questions.
Get your AI Visibility Score, free
12 real buying questions for your trade and market, across three assistants, three runs each, your X/12, the screenshots, and the first fix worth making. No sales call, and the findings are yours to keep.
Get my free Visibility Check