AI Visibility Score: What It Means and How to Check Yours
An AI visibility score is, in most trackers, the percentage of tracked AI answers that mention your brand. It only means something next to competitors measured on the same prompts, engines and formula, and because it comes from a sample of answers, it should be quoted as “x of n answers”, not as a bare number.
By Harjot Chopra, Co-Founder, PageLens.ai · Updated 16 Nov 2026

Key takeaways
- In most trackers, an AI visibility score is the percentage of tracked AI answers that mention your brand: answers that mention you ÷ all answers × 100. Some tools blend in citations, sentiment or position, and page-level graders measure something else, so check which kind of score you are looking at.
- The formula, the engine mix, the prompt wording and the run all move the score, so two tools can give the same brand different numbers. Compare only scores built the same way, and quote any score as “x of n answers”.
- No standard “good” score exists. Good means level with or ahead of your top competitors on the same prompts and engines.
- To check yours, fill in the sheet below (20 prompts with no brand names, run on 3 engines, 60 answers). Report each engine as well as the pooled score (all engines combined), treat pooled gaps under about 20 points as too close to call, and treat engine gaps under about 30 points as indicative only.
In this article
- 1What is an AI visibility score and how is it calculated?
- 2Why does the same brand get different AI visibility scores?
- 3What is a good AI visibility score?
- 4How do you check your AI visibility score yourself?
- 5When should you track your AI visibility score continuously?
- 6What should you do with your AI visibility score?
- 7Frequently asked questions
The same 80 AI answers can produce a visibility score of 45 or 25, depending only on how you count them. Count every answer that mentions your brand and you get 45. Give a first-place mention more weight than a passing one and you get 25. The working is shown below.
That is why a score from a free checker cannot, by itself, tell you whether you are doing well. Every score is a measurement someone designed: which questions were asked, on which AI engines, how many times, and what counts as a mention. Change any one of those and the number can move by many points, and more than 20 vendors now sell these scores, each built differently.
This page gives you the formula, the evidence on what moves the number, and a 60-answer sheet you can fill in over an afternoon to see where you stand against the competitors who matter. It also shows how to tell when two scores are too close to call, so you do not rebuild your content plan around noise. It is written for marketing and SEO leads who already have a number from a checker and need to know what to do with it.
What is an AI visibility score and how is it calculated?
An AI visibility score is, in most trackers, the percentage of tracked AI answers that mention your brand. Calculate it as answers that mention your brand ÷ all answers × 100, so 30 mentions in 100 answers is a score of 30. This page uses that definition, which the IAB calls Mention Rate in its Measuring Visibility in the AI Era framework (August 2026).
An "AI engine" here means a chatbot or AI search product such as ChatGPT or Perplexity. A mention means your brand name appears in the answer text. A citation means the answer links to your site. Citation rate is a separate measure: answers that cite your site ÷ all answers × 100. A brand can be named without a link, and linked without being named, so one number cannot stand in for the other. A pooled score combines the answers from all engines into one percentage.
Some scores weight a mention by how prominent it is. This page calls the weighted version the points score and the unweighted percentage the mention rate. They are different formulas, not better and worse ones, and the next section shows how far apart the two can land.
Not every score is built this way. Some vendors blend mentions with citations, sentiment, position and share of voice into one composite. HubSpot's AI visibility score guide, for example, lists platform coverage, mention frequency, citations, sentiment, consistency and share of voice as inputs to a composite, while its own AEO tool reports brand visibility as the plain share of tracked answers that mention the brand.
Page-level graders measure something else again: our free AI Visibility Score tool grades a single page from 0 to 100 on how readable and citable it is to AI engines, not on how often AI answers name your brand. Before you compare two numbers, check which of these three they are.
A visibility score is also easy to confuse with its neighbours, because each one divides by something different.
| Metric | Formula | What it tells you |
|---|---|---|
| AI visibility score (mention rate) | Answers that mention you ÷ all answers × 100 | How often a buyer sees your name |
| Citation rate | Answers that cite your site ÷ all answers × 100 | How often your site is credited as a source |
| AI share of voice | Your mentions ÷ all brand mentions × 100 | Your slice against competitors when brands are named |
| Sentiment | Share of your mentions labelled positive, neutral or negative | How the answers describe you |
The IAB's framework defines share of voice as brand mentions as a proportion of all brand mentions within a defined competitive category. The score counts answers that name no brand at all, while share of voice ignores them. If a rival drops out of many answers, your share of voice can rise while your score stays flat, so report the two side by side.
Why does the same brand get different AI visibility scores?
The same brand gets different scores because four choices sit under every score: the formula, the engines, the wording of the prompts and the number of runs. Each one has measurable effects.
The IAB's Measuring Visibility in the AI Era (August 2026) reports that more than 20 companies now sell AI visibility measurement, each with its own methods, prompt libraries, platform coverage and scoring rubrics, and warns that two providers measuring the same brand in the same category can return materially different results. The four choices below show how that happens.
The formula. Take an illustrative log of 80 answers: your brand is the primary recommendation in 10, a secondary mention in 12, a passing mention in 14, and absent from 44. This is arithmetic we worked through to show the mechanism, not a measurement.

| Outcome | Answers | Points (5 / 3 / 1 / 0) |
|---|---|---|
| Primary recommendation | 10 | 50 |
| Secondary mention | 12 | 36 |
| Passing mention | 14 | 14 |
| Absent | 44 | 0 |
| Total | 80 | 100 |
The plain mention rate is (10 + 12 + 14) ÷ 80 × 100 = 36 ÷ 80 × 100 = 45. The points score is 100 ÷ (5 × 80) × 100 = 100 ÷ 400 × 100 = 25. Same answers, two defensible scores.
The engines. In Citation Research Study 003, our own lab's study with fieldwork on 5 October 2026, we asked four engines the same software-buying questions from India without naming India. The engines were ChatGPT, Perplexity, Gemini and Google AI Mode (Google's chat-style search mode). We counted the share of product mentions that were Indian software:
| Engine | Indian products named | Product mentions | Share |
|---|---|---|---|
| Google AI Mode | 30 | 117 | 26% |
| ChatGPT (search on) | 21 | 119 | 18% |
| Perplexity | 8 | 88 | 9% |
| Gemini | 10 | 120 | 8% |
| All four (our sum) | 69 | 444 | 16% (15.5% before rounding) |
The table shows that engines differ in what they surface for the same questions. In this study the Indian share ranged from 8% to 26% by engine. It is a share of product names, not a brand mention rate, and it covers Indian small-business software. We have not measured a brand mention rate by engine, but we expect the same spread, so report each engine separately.
The wording. In the same study, 16% of 444 product mentions were Indian when the question did not mention India. When the question named or hinted at India, 65% of 906 mentions in 192 answers were Indian. The brand set did not change; the question did. Study 003 tested India cues, not brand names, but the same logic applies to brands.
Searchable's AI visibility score guide (30 September 2026) reports, from the 3,736 brands tracked on its own platform in the 30 days to 17 September 2026, that the median brand was named in 93% of answers to branded prompts and 7% of answers to unbranded ones. That is one vendor's own data, so treat it as an illustration of the size of the effect. Mixing the two lets the prompt list, not your visibility, set the score, so keep brand names out of the 20 prompts below.
The run. In a SparkToro study fielded in November and December 2025, 600 volunteers ran 2,961 prompts across ChatGPT, Claude and Google AI. The same brand list came back less than 1 time in 100, and the same list in the same order less than 1 time in 1,000 (reported by Search Engine Journal on 30 January 2026). Our own Study 001 points the same way: across 800 answers on four engines, only 20% of 400 repeat pairs named exactly the same providers. Visibility percentages across many prompts and repeated runs are steadier than any single answer, which is why a score is a sample and not a fact.
What is a good AI visibility score?
A good AI visibility score is one that is level with or ahead of your top competitors, measured on the same prompts, engines and formula. No fixed number is "good". In SparkToro's headphone test (994 responses), leading brands appeared in 55 to 77% of answers. That is one famous-brand category, so our inference is that a "good" threshold varies by category.
Averages quoted without a prompt list, engine list and formula tell you nothing about your market, so ignore them.
Here is how we would act on the gap between your pooled mention rate and your best competitor's, from the same 60 answers. The thresholds are our model, not an official figure. A single rate at 50% on 60 answers carries about ±13 points (1.96 × √(0.5 × 0.5 ÷ 60) = 12.7). A gap between two rates carries more: 1.96 × √(2 × 0.25 ÷ 60) ≈ 18 points at 50%, assuming independent samples. We therefore set the "too close to call" line at about 20 points. Split across 3 engines, each engine has 20 answers and one rate carries about ±22 points (1.96 × √(0.25 ÷ 20) = 21.9), so engine gaps below about 30 points are indicative only.
| Your gap to the top competitor | What it means | What we would do |
|---|---|---|
| Behind by more than 20 points | You are being skipped in answers your buyers read | List the prompts where the competitor is named and you are absent. Use engine gaps to choose where to look first, not as proof |
| Within 20 points either way | Too close to call | Repeat the sheet in 4 to 6 weeks before changing the plan |
| Ahead by more than 20 points | You lead on this prompt set | Check each engine; a lead can hide an engine where you are absent, though engine gaps under about 30 points are indicative only |
For the margin-of-error working behind these lines, see our sampling sheet for measuring AI search visibility and how many answers make an AI citation number real.
How do you check your AI visibility score yourself?
Check it in two steps: get a free baseline, then build a 60-answer sheet you control. The baseline is quick but is one read; the sheet is slower but lets you see how the score was made.
Step 1: baseline. Our Free AI Visibility Audit takes your domain, writes 10 buyer prompts from your site and runs them across AI engines including ChatGPT, Gemini and Perplexity. It returns a 0 to 100 visibility score, a sentiment split, the words the AI models use about your brand and the competing domains they recommend, in about 30 seconds with no signup and no credit card. Use it as a first read, not a benchmark: a tool that writes the prompts for you can include some that name your brand, so run its prompts through the four checks below before you trust the number.
Step 2: the sheet.
-
Write 20 prompts a buyer in your category would type: 5 category questions, 5 comparison questions, 5 "which should I choose" questions and 5 use-case questions. Check each one: if the prompt contains your brand name, rewrite it. Pass: none of the 20 prompts contains your brand name. Fail: any prompt names you; rewrite it. Our guide to choosing AI prompts worth tracking covers this step in depth.
-
Run each prompt once on 3 engines, signed out, from the same location, each in a fresh session. That is 60 answers.
-
Log one row per answer: prompt, engine, date, your position (P primary, S secondary, M passing, A absent), competitors named, and whether your domain was cited.
-
Compute the scores with these spreadsheet formulas, assuming your position codes are in D2:D61:
Mention rate:
=(COUNTIF(D2:D61,"P")+COUNTIF(D2:D61,"S")+COUNTIF(D2:D61,"M"))/COUNTA(D2:D61)*100Points score:
=(5*COUNTIF(D2:D61,"P")+3*COUNTIF(D2:D61,"S")+COUNTIF(D2:D61,"M"))/(5*COUNTA(D2:D61))*100 -
Repeat the same calculation for each competitor, for each engine on its own, and for citation rate.
-
Write the context beside every number: "x of n answers, which engines, prompt set version, date, formula".
Before you trust any figure, run it against these checks:
- Denominator. Pass: the figure reads "x of n". Fail: only a percentage.
- Engines. Pass: the engines are named and a per-engine split is shown. Fail: one pooled number.
- Prompts. Pass: you can see the prompts, and none names your brand. Fail: the prompts are hidden or any names your brand.
- Formula. Pass: the same formula for both periods. Fail: the method changed between readings.

When should you track your AI visibility score continuously?
Track continuously once you need to see a trend, because one sheet is a single reading and the score moves with each run. The sheet suits a first look and a repeat every 4 to 6 weeks while you are changing content, then quarterly once things are stable; a fixed prompt set run repeatedly suits ongoing decisions.
On our paid plans we re-run a fixed prompt set across your tracked engines. Launch and Growth each include 100 visibility prompts a day across 10 categories, with up to 10 competitors each, for one country (more countries are add-ons). Engines by plan:
- Launch tracks ChatGPT, Google AI Mode and Perplexity.
- Growth adds Gemini and Grok.
- Enterprise adds Claude and Copilot, for seven engines.
As of November 2026, Launch costs $299 a month, Growth $699 a month and Enterprise starts at $1,499 a month; see current pricing. To see how your prompt set and competitors would be set up, Book a demo.
What should you do with your AI visibility score?
Your score is the start of a diagnosis, not the verdict, so open the answers behind it. In the next 48 hours, write your 20 prompts, confirm that none names your brand, run them on three engines, and mark every answer where a competitor is named and you are not. For each one, note which pages the engine cited: those are the pages to study first, and our guide to getting cited by AI shows how to approach them.
Report the result as "x of n answers, on these engines and prompts", never as a bare number. If you want a quick outside read to compare your sheet against, run the Free AI Visibility Audit, which takes about 30 seconds with no signup.
Track your AI visibility against competitors
Frequently asked questions
Sources
- 1.Measuring Visibility in the AI Era (IAB, August 2026)
- 2.New research: AIs are highly inconsistent when recommending brands or products
- 3.AI Recommendations Change With Nearly Every Query: SparkToro
- 4.Study 003: What AI recommends to Indian small businesses buying software
- 5.Study 001: AI coaching recommendations in India
- 6.Study 008: Asked from India, four engines cited American sites 3.4 times as often as Indian ones
- 7.AI visibility score: How to summarize your AI visibility, HubSpot, updated 23 October 2026
- 8.What Is an AI Visibility Score? How to Calculate It, Searchable, 30 September 2026


