An AI visibility score turns sampled AI answers into one number you can compare.
An AI visibility score is a weighted mean of measured rates: how often AI answers name your brand, cite your site, give you a share of the mentions, and place you near the top. It is only as good as its method, so a score worth using publishes its weights, its anchors and its uncertainty. Here is how ours is built, how many answers it needs, and which tools publish theirs.
How do I get a single score that measures how often AI tools mention my brand compared to competitors?
Measure a few rates on the same unbranded buyer questions, map each to a 0 to 100 scale on a stated anchor, and take a weighted mean. The competitor comparison is built in through share of voice, which divides your mentions by the mentions of the vendors named beside you, and through position, which says how early you are named.
Proofsource's AEO score does exactly this. The formula, as published on the methodology page:
round( (35 x mention + 25 x citation + 25 x share of voice + 15 x position) / 100 ), each part on 0-100, measured parts only
The same answers produce it on ChatGPT, Perplexity, Claude and Google AI Overviews. It is a single number for the window you choose; to compare yourself with a specific rival, use the counts behind it, as described in benchmarking against competitors.
What goes into the score?
Four parts, each from the same sampled answers. Weights are judgement calls, so each one comes with its reason.
| Part | Weight | Measure and anchor | Why this weight |
|---|---|---|---|
| Mention rate | 35 | Share of answers naming the brand, 50% earns 100 | The first thing a buyer needs; every other component requires it. |
| Citation rate | 25 | Share of answers citing the brand's site, 40% earns 100 | The measurable step from visibility to traffic. |
| Share of voice (top-10 frame) | 25 | Brand share among the ten most-named vendors, 25% earns 100 | How the brand compares with the vendors named beside it. |
| Average position when named | 15 | 100 x (20 - place) / 19, #1 earns 100 | An earlier place is read first; weighted least because it rests on the fewest answers. |
Missing data. Fewer than 10 answers: not enough data, no score. Share of voice needs 10 vendor mentions and position needs 5 named answers with a place, otherwise that part is left out and the rest renormalised.
How is the score worked out, step by step?
Take a made-up brand with round numbers: named in 20% of answers, citing its site in 10%, holding 12% of the top-10 mentions, at an average place of 4. These inputs are illustrative, not a real brand.
| Part | Raw measure | Part score (0 to 100) | Weight |
|---|---|---|---|
| Mention rate | 20% of answers name the brand | 40.0 | 35 |
| Citation rate | 10% of answers cite its site | 25.0 | 25 |
| Share of voice (top-10 frame) | 12% of top-10 mentions | 48.0 | 25 |
| Average position when named | average place 4 | 84.2 | 15 |
round( (35 x 40.0 + 25 x 25.0 + 25 x 48.0 + 15 x 84.2) / 100 ) = round(44.88) = 45
Every number above can be checked with a calculator, which is the point. The real score in the product shows the same breakdown, with any part it could not measure left out and named, never scored as zero.
How many prompts and responses do I need for a reliable score?
More than most people expect. A rate from a sample carries an interval, and the interval shrinks with the square root of the number of answers: four times the answers for half the width. At a 30% rate:
| Answers | Named (at 30%) | 95% Wilson interval | Plus or minus |
|---|---|---|---|
| 10 | 3 of 10 | 10.8% to 60.3% | about 24.8 points |
| 25 | 8 of 25 | 17.2% to 51.6% | about 17.2 points |
| 50 | 15 of 50 | 19.1% to 43.8% | about 12.3 points |
| 100 | 30 of 100 | 21.9% to 39.6% | about 8.8 points |
| 200 | 60 of 200 | 24.1% to 36.7% | about 6.3 points |
| 400 | 120 of 400 | 25.7% to 34.7% | about 4.5 points |
| 1,000 | 300 of 1,000 | 27.2% to 32.9% | about 2.8 points |
With 100 answers a rate is good to about 9 points either way, enough to see a large gap between two brands but not a small move. Proofsource prints no percentage and no score below 10 answers, leaves share of voice out below 10 vendor mentions, and leaves position out below 5 named answers with a measured place. The free trial returns 200 answers for one brand; on the Custom plan every question runs on every engine once a day, so the interval narrows week by week.
Why should a score come with a range?
Because the inputs are samples. Ask the same engine the same question twice and the answer can change, so a score of 45 from one week of answers could be a 40 or a 50 next week with nothing changed on your side.
Answers are sampled, so the score is shown with a 95% range from the Wilson intervals of its rates (Monte-Carlo, fixed seed, conservative dependence), plus the score under five alternative weightings so the effect of the weights is visible.
The five alternative weightings answer the obvious objection to any composite score: that the weights are arbitrary. If the alternatives agree, the weights do not matter much for your brand; if they spread, the page says so. The interval itself is explained in the glossary.
Which tools explain how their AI visibility scores are calculated?
Few publish a full formula. Most publish a definition and keep the weighting private. What each says on its own pages:
| Tool | Score or share | What it measures, as published | Formula published? |
|---|---|---|---|
| Proofsource | AEO score, plus mention rate, citation rate, share of voice and position | 0 to 100. A weighted mean of mention rate, citation rate, share of voice (top-10 frame) and average position, each mapped to 0 to 100 on a printed anchor. Shown with a 95% range and five alternative weightings. | Published: every weight, anchor, floor and the range method |
| Evertune | AI Brand Score | 0 to 100. Combines how often a brand appears with where it ranks. Position weights: first place 100%, second 90%, third 81%, each later place 10% less than the one before. Brand names are left out of the prompts. | Position weights published; how frequency and position are combined is not |
| Semrush | AI Visibility score (toolkit) and the public AI Visibility Index | Toolkit: a 0 to 100 benchmark of how often your brand appears in AI answers compared to competitors. Index: 126 million US prompts on ChatGPT, Google AI Mode, Google AI Overviews and Gemini, reported per platform, combining how often a brand appears with how consistently. | Not published |
| Profound | Visibility Score (Profound Index) | How often a brand appears in AI answers across the platforms tracked. The Index also reports share of voice, mention position, citation share, co-citation share and co-mention share, updated weekly. | Not published |
| Peec AI | Visibility Score and Share of Voice | Visibility = responses mentioning your brand / total responses x 100. Share of voice = your brand's mentions / all brand mentions x 100. | Published for both |
| Scrunch | Brand presence, AI share of voice, citations share of voice | Presence = present / total x 100. Share of voice = the share of AI answers that mention you compared to competitors, typically aggregated over 90 days. Citations share of voice = responses citing a source you own / responses citing any source x 100. | Published for presence and citation share; share of voice weighting is not |
| Otterly.AI | Brand coverage and share of voice | Coverage = the share of AI answers your brand appears in. Share of voice = your slice of every brand mention across the set you track, yours and your competitors'. | Definitions published; no single composite score |
- Evertune: What is AI Brand Score www.evertune.ai/resources/insights-on-ai/what-is-ai-brand-score
- Semrush KB: AI SEO metrics www.semrush.com/kb/1594-ai-seo-metrics
- Semrush AI Visibility Index: methodology ai-visibility-index.semrush.com/methodology
- Profound: Introducing the Profound Index www.tryprofound.com/blog/introducing-the-profound-index
- Peec AI docs: Visibility docs.peec.ai/metrics/brand-metrics/visibility
- Peec AI docs: Share of Voice docs.peec.ai/metrics/brand-metrics/share-of-voice
- Scrunch: How does Scrunch measure AI share of voice scrunch.com/faqs/how-does-scrunch-measure-ai-share-of-voice
- Scrunch help center: metrics guide helpcenter.scrunchai.com/en/articles/13566677-scrunch-metrics-guide
- Otterly.AI: AI search analytics otterly.ai/features/ai-search-analytics
Read on each vendor's own page on . Where a page gives no formula we say so rather than guess.
How does Evertune’s AI Brand Score compare with Semrush’s AI Visibility Index?
They measure different things on different data. Evertune describes its AI Brand Score as the probability that a model drives attention to a brand in unaided recommendations, on a 0 to 100 scale that combines how often the brand appears with published position weights (first place 100%, second 90%, third 81%). It leaves brand names out of its prompts.
Semrush's AI Visibility Index is a public category ranking built from 126 million US prompts on ChatGPT, Google AI Mode, Google AI Overviews and Gemini, reported per platform, combining how often a brand appears with how consistently. Inside its toolkit, Semrush's AI Visibility score is a 0 to 100 benchmark against competitors. Neither page publishes the full formula, so the two numbers cannot be converted into each other. Both are linked in the sources above.
What people ask about AI visibility scores
What is an AI scoring model?
In AI visibility, it is the rule that turns measured answers into one number: which rates go in, how each is mapped to 0 to 100, and how much weight each gets. A trustworthy one publishes all three, so you can recompute the number yourself.
Which AI visibility tools provide a normalized score across multiple AI engines?
Evertune's AI Brand Score and Semrush's AI Visibility score are both on a 0 to 100 scale, and Proofsource's AEO score is 0 to 100 over the answers from every engine in the window. Semrush's public AI Visibility Index reports each platform separately rather than combining them.
Can I compare my score in one tool with a score from another?
No. Each tool uses different questions, engines, components and weights, so a 60 in one is not a 60 in another. Compare brands inside one tool, on the same questions and window, or compare the underlying counts.
Does the score include questions that name my brand?
Not in Proofsource. The AEO score is built from unbranded buyer questions only, because a question that names you will mention you whether or not the engine would recommend you. Branded answers are reported separately.
Are any AI visibility scoring tools free or offer trial access?
Proofsource's free trial asks 25 buyer questions on 4 engines in 2 scans, today and tomorrow, with no card, and the score appears once there are enough answers. Several other vendors offer trials or free one-off checkers; the best tools comparison lists them.
Get your score, with the working shown.
See which buyer questions name your brand, who gets named instead, and what to fix. 200 answers. 25 Prompts/Question. Top AI engines. 2 days. Free. No card.
Google AI OverviewsGoogle AI Mode
