What is an AI visibility score?
An AI visibility score is a single 0–100 figure for how present a brand is in AI-generated answers. A credible one combines how often assistants actually name and recommend you with how readable your site is to them — and it only means something once the same score has been measured for your competitors, not just for you.
Updated 21 July 2026
We publish a score — the AVI Score — so read this knowing that, and knowing that a score you cannot take apart is a score you cannot defend to anyone who asks.
What the number is actually made of
Our AVI Score runs 0 to 100 and combines two things, because they answer two different questions and a brand can pass one while failing the other.
Outcome (60%) — do assistants name you? We ask ChatGPT, Gemini, Claude and Perplexity the questions your buyers actually ask, many times each, and count how often you are named, how often you are recommended, and how often you are cited as a source. This is the half that decides whether you exist in an answer, so it carries most of the weight.
Readiness (40%) — can assistants read you? A crawl scores the technical signals an AI system leans on to understand and quote you: server-rendered content, structured data, a clearly-defined entity, crawler access. This half is cheap to check, and you can scan it free.
The split is the useful part. A flawless, perfectly readable site that no assistant ever names scores high on readiness and low on outcome — and the honest reading of that gap is that your website is not the problem. Most tools give you one blended figure and hide which half is failing.
Why a single number at all
Because the places a score has to travel — a board paper, an investor deck, a claim on your own homepage, a comparison with a named competitor — need a figure a non-specialist can hold in their head. A dashboard of twelve metrics does not survive that journey.
The cost of compressing to one number is that it hides detail, so the only honest way to publish one is to make it decompose on demand (into outcome and readiness, and those into their inputs) and to rest it on a method anyone can read. A score that cannot be opened up is an opinion with a decimal point.
Why the score is meaningless on its own
Here is the part most "AI visibility score" products skip. A rate is not a position.
If your score is 8 out of 100, that sounds like a failure — until you learn it was the highest figure in your category. A number describing you alone tells you how often you were named; it cannot tell you whether that is good, because good is defined by the field. The metric that answers "where do I stand" is share of voice, and share of voice is a fraction whose denominator is everyone else's mentions — which someone has to go and measure.
That is the difference between a score and an index, and it is why we argue you should not trust a private number without a measured field behind it. What separates an index from a dashboard goes through it in full.
How to read a score honestly
Three things to hold on to, all learned the hard way from our own readings.
It is a rate, not a certainty. Assistants answer the same question differently each time, so every figure carries a range. One of our recent readings put a brand's mention rate at 10.7% — with the true value somewhere between 7.5% and 15.1%, across 252 sampled answers. Anyone quoting a bare percentage is quoting the midpoint of a range they have not shown you.
Movement inside that range is noise. A score that drifts from 8 to 9 and back is dice, and it will be sold to you as progress. We refuse to draw a trend unless the movement clears the interval.
A score can move because the method moved. If the question set or the set of assistants changes and your number shifts, that is the measurement moving, not your brand. A serious score versions its method and breaks the trend line rather than quietly redrawing it — when we added a fourth assistant, readings taken before it became non-comparable to readings taken after, on purpose.
What a score cannot do
- Guarantee a recommendation. It is a measurement of the present, not a purchase. Nobody can promise a mention, and anyone who does is selling certainty that does not exist.
- Predict tomorrow. A reading is a repeatable measurement, not a forecast.
- Stand in for a measurement you did not take. We know because we once let seeded demonstration numbers reach a public ranking. The rule that replaced it is absolute: a score is either measured or it is not published. A real, low score beats an invented, flattering one every time.
You can see the readable half of your own score in a free scan, and the full picture — outcome as well as readiness, against your category — is what the index publishes.