What is an AI visibility score?

What is an AI visibility score?

An AI visibility score is a 0–100 measure of how often AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Bing Copilot — name your brand when someone asks buying questions in your category. It is not a Google ranking, and it doesn't measure how much an AI "knows" about you. It measures one thing: when a potential customer asks, are you in the answer?

There's no industry standard behind the phrase. Every tool that offers a score defines its own method, which makes the method the thing to check before you trust any number — including ours.

How is the score calculated?

Any honest score has three parts: a set of questions, a way of collecting answers, and a scale for what counts. For our free audit, that means:

1. Six buyer questions, generated from your site — the category as a stranger would phrase it, with your geography included.
2. Live web evidence per question — the same retrieval the answer engines run, so the score can't be flattered by a model's stale memory of you.
3. A weighted verdict per question — being recommended outscores being cited, which outscores being merely named. The distinctions are covered in guide № 01.

Tools that instead ask a model directly "have you heard of X?" measure training memory, and training memory is stale by definition. Check which kind of score you're being sold.

What's a good score?

Starting below 40 is normal, not a verdict. The score is an inventory of today's evidence about you on the web, and evidence can be built.

How is it different from a Google ranking?

A ranking is a position: your page is third for a query. A visibility score is a presence: your brand appears in the composed answer, or it doesn't — there is no third place in a paragraph. The engines share crawling machinery with search, but they reward different things: a ranking rewards the best page, an answer rewards the most quotable, best-corroborated brand. Page one on Google with a weak score is common, and it means your rankings aren't translating into being the answer.

Why does the score change week to week?

Because the inputs do. Answer engines re-retrieve their sources constantly: a roundup updates, a competitor's review count passes yours, a directory reshuffles, and this week's answer quietly differs from last week's. Single-digit wobble means nothing. What matters is the trend over months and any step change after you fix something — which is the reason to score monthly with the same method rather than once.

How do you improve it?

Improving the score means changing the evidence the engines read, and the wrong fix wastes months. The five failure modes and their fixes have their own guide: why ChatGPT doesn't mention your brand. The short version: get named in the sources engines cite, give them quotable sentences on your own site, and make sure the crawlers can read you at all.