What does your AI visibility score mean, and what doesn't it tell you?

Your AI visibility score is a number from 0 to 100. In a Seen in Search report, it shows how often AI assistants named your business when asked your buyers' questions, how high you came in their lists, whether they linked to your website and how warmly they described you.

A score of 60 or more puts you in the Visible band, where assistants name you, though not always first. 80 or more is Category Leader, where answers name you first. The number describes the answers given on the day they were asked. It makes no promise about tomorrow's.

What goes into the score?

Five measurements, each scored from 0 to 100 and combined with fixed weights. We publish the weights on our methodology page, and they are the same numbers the scoring code uses.

Presence carries the most weight, at 35%.

When a report asks more than one AI assistant, each assistant's answers also carry a share of the total. The methodology page lists those shares.

How is the number worked out?

Here is the arithmetic for the illustrative example on our homepage, a made-up firm called Acme. Its five measurements came out at 71, 58, 44, 55 and 78.

Acme's weakest measurement is share of voice. Its three rivals were named five times between them, against Acme's four. Its strongest is sentiment, because the answers that named Acme described it well.

A language model reads each answer and notes who was named, in what order, with which links and in what tone. It does not set the number. The score is computed in code from the fixed weights, so the same answers always produce the same score.

What do the five bands mean?

Read the band first and the exact number second. These are the five bands on our methodology page:

Where you need to be depends on your rivals. In the illustrative example, a rival called Northwind scored 74 and was usually named ahead of Acme. Both firms sit in the Visible band, yet Acme's 61 left it 13 points behind.

A move across the line between two bands does not by itself mean buyers see you differently. Look at what changed in the answers before you read much into it.

How do you check your AI visibility score?

Start by hand. Type a question a buyer would ask while choosing a firm like yours into an AI assistant. Note the date, every business named, the order they came in and any links shown.

Here is one we ran. On 14 November 2026 at 10:46 UTC, we typed "best chartered accountant firm for startups in bengaluru" into Google, in English, set to India. Its AI Overview named eight firms. Five got a line of description each, and the other three shared one sentence at the end. The first three sources the answer showed were firms' own websites.

In that one answer, being among the eight counts towards presence, and being in the first five, with a line of your own, towards prominence. The seven other firms named beside you set your share of voice. A link to your own website would count as a citation, and the words used about you as sentiment.

One answer is one sample, though. Ask again and the list can change, as the research further down shows.

Our report does this for you with a fixed method. You sign up with a business email and a one-time code, with no card and no password. We read your homepage first and write the buyer questions from what it says you sell, so the questions are about your category. The score then uses the published weights. The methodology page lists how many questions and AI assistants each report uses.

When a report finishes, it is private. You choose whether to publish it, and can take it down again. Reports are never offered to search engines for indexing.

What doesn't the score tell you?

Your Google ranking is a separate measure. The score reads AI answers. Where you rank in Google's ordinary list of links is a different matter, and a page can rank well there while missing from the answer a buyer reads first.

The score makes no promise. Our methodology page calls it an observational score based on samples of AI answers, and says it is not a guarantee. It does not forecast enquiries or sales.

Only the questions asked count. The score comes from the buyer questions in your report, written from what your homepage says you sell. Ask different questions and different firms can come up.

It cannot see why. The score records what the answers said. Beyond the links an answer showed, it cannot tell which page or review an assistant learned from.

Why can a fresh report give a different score?

AI assistants write a fresh answer each time they are asked, and the lists in those answers change a lot.

SparkToro, an audience research company, measured this in research published on 27 January 2026. Its 600 volunteers ran 12 different prompts through three AI tools a combined 2,961 times, with each prompt run 60 to 100 times.

• For ChatGPT and Google's AI answers, SparkToro put the chance of any two answers giving the same list of brands at under 1 in 100. The same list in the same order was more like 1 in 1,000.
• SparkToro concluded that how often a brand appears across many runs of many prompts is a reasonable measure. It was much harsher on position, calling any tool that reports a "ranking position in AI" nonsense.
• Its advice for anyone who wants to know what an AI recommends is to ask the same question over and over, usually at least 60 to 100 times, and average the results.

For your score, this means one report is one sample. It asks each question once per AI assistant, on one day, while SparkToro's measure came from dozens of runs of each prompt. Treat presence in a single report as a first reading. Our score also gives position some weight, inside prominence at 20%, and SparkToro's findings are a good reason to read that measurement last.

How do you increase your AI visibility score?

Start with presence, because it carries the most weight. If the answers already name you most of the time, move down the list.

1. Presence. If answers leave you out, work on that first. Check that your homepage says plainly what you sell and who you sell it to. Our questions are written from it, and a buyer skimming it should get the same picture.
2. Citations. In our Google check, the first three sources the answer showed were firms' own websites. If answers name you but link to other sites, make sure your own pages describe the work you do in plain words, with figures you can back with a named source.
3. Share of voice. This depends on how often rivals are named as well as you, so it can move when nothing on your site changes. In the illustrative example, it was Acme's weakest measurement.
4. Prominence and sentiment. These describe the answers that already mention you, so they come last.

Then check again, asking the same buyer questions the same way, so the two sets of answers can be compared.