An AI visibility score is the share of AI assistant answers, to a set of questions your customers might ask, that mention your business. Ask ChatGPT, Perplexity, Gemini, Claude or Google's AI features the same 20 questions, count how many answers name you, and that percentage is your score. It tells you whether AI is recommending you at all, and over repeated checks whether that is getting better or worse.
Key takeaways
- The score is a percentage: answers that mention you, divided by all the answers collected, for a fixed set of questions and assistants.
- It is only meaningful with its method: which questions, which assistants, how many runs and over what period.
- Supporting metrics (share of voice, average position, citation rate, sentiment) explain why the score is what it is.
- A single check varies because AI answers vary. The trend across repeated checks is what matters.
- There is no published industry benchmark for a "good" score. Most small businesses start near zero, and being named consistently in a few answers is the realistic first milestone.
How is an AI visibility score calculated?
The score comes from three choices (questions, assistants, repeats) and one simple count: how many of the collected answers mention you.
- A set of questions. These stand in for what customers ask an assistant. For a plumber in Leeds, examples might be "who is a reliable emergency plumber in Leeds?", "how much does a boiler service cost in Leeds?" and "best plumber near Headingley". These are illustrative examples, not measured search volumes. What matters is that the set is fixed, so each check compares like with like.
- A set of assistants. Each finds businesses differently. ChatGPT's search features surface websites through OpenAI's OAI-SearchBot crawler (OpenAI, 2026); Perplexity's PerplexityBot is "designed to surface and link websites in search results on Perplexity" (Perplexity, 2026); Google's AI features rely "on our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index" (Google Search Central, 2026). Because the sources differ, a score per assistant is more useful than one blended number. Our guide on how AI assistants choose businesses covers each platform.
- Repeated runs. Each question is asked on a schedule (weekly, fortnightly or monthly), and often more than once per run, because answers differ from one attempt to the next.
For each answer, a tool records four things: whether you were mentioned (the yes/no that feeds the score); your position if several businesses are listed; whether your website was cited as a source or you were named from a directory, review site or article; and who else was named.
The score itself is:
answers that mention you ÷ all answers collected × 100
A worked example. You track 20 questions across 2 assistants, each asked once a fortnight: 40 answers per check. If 6 name your business, your score for that check is 15. If a later check gives 10 out of 40, it is 25.
Tools differ in the details: some weight by position, some count only the first few businesses named, some blend assistants into one figure. None of that is wrong, but a 30 from one tool and a 30 from another are not the same thing. Read the method before comparing.
What are the related AI visibility metrics?
The score answers "am I named?". Four supporting metrics answer "compared with whom, how prominently, on what evidence, and said how?".
| Metric | What it measures | Worked example (40 answers per check) |
|---|---|---|
| AI visibility score | Share of all answers that mention you | 10 of 40 answers name you: score 25 |
| Share of voice | Your mentions as a share of all mentions of you and your tracked competitors | You: 10 mentions. Competitor A: 15. Competitor B: 10. Competitor C: 5. Total 40, so your share of voice is 25 per cent and A leads with 37.5 |
| Average position | Where you appear in answers that list several businesses | Named 1st, 2nd, 2nd, 4th, 1st: average position 2.0 |
| Citation rate | Share of answers that link to your website as a source, rather than naming you from a third-party page | 4 of the 40 answers cite your site: citation rate 10 per cent, even though you were named in 10 |
| Sentiment | Whether the mention is positive, neutral or negative ("highly rated", "mixed reviews", "closed down") | 8 positive, 2 neutral, 0 negative |
Reading them together:
- Score high, citation rate low means assistants know you from directories, reviews and articles rather than from your site. Google confirms its AI features "can show what's being said about products and services across the web, including in blogs, videos, and forum discussions" (Google Search Central, 2026). Being named is good, but it leaves your prices and services to be described by someone else's page.
- Score steady, share of voice falling means a competitor is being added to answers more often than you. The headline score hides this; share of voice shows it. See our guide to AI competitor analysis.
- Average position matters most for list-style answers ("the top five accountants in Bristol"), where fifth gets far less attention than first.
- Sentiment rarely moves quickly, but a negative pattern usually traces back to one review or news page that assistants keep finding.
Why does a single AI visibility check vary?
Because AI answers are not fixed. The same question, asked twice, can produce different businesses, so one check is a snapshot and the trend over several checks is the signal.
Three documented reasons:
- The question is rewritten and searched live. Google describes "issuing multiple related searches across subtopics and data sources" to build an AI Overview or AI Mode response (Google Search Central, 2025). Which pages come back changes as pages are added, updated and re-ranked.
- Location changes local results. Google says local results are "mainly based on relevance, distance, and popularity", where "Distance refers to how far each business is from the customer who's searching" (Google Business Profile Help, 2026). A check run from an office in Manchester is not the same as a customer asking from Salford. AI features also differ by country: Google lists where AI Overviews are shown, and the UK is among them (Google Search Help, 2026).
- The text is generated, not retrieved. Even with the same sources, the assistant composes a fresh answer each time and may pick a different three businesses from the same ten candidates. Google's own help page advises users to "Ask multiple versions of your question to get the best answers" (Google Search Help, 2026), which is the same problem seen from the other side.
What this means for measurement:
- Fix the question set and leave it alone between checks, or you cannot tell whether the score moved because of you or because of the questions.
- Ask more than once. Twenty questions asked once is 20 data points; the same set across three assistants, fortnightly for two months, is 240. Variation averages out.
- Watch the direction, not the decimal. A move from 15 to 18 in one check is noise. A move from 10 to 25 over three months, across all assistants, is a result.
- Compare like with like. Same questions, same assistants, same location setting, same tool.
What is a good AI visibility score?
There is no published industry benchmark, and we are not going to invent one. What we can say is where most small businesses start and what a sensible first target looks like.
Most small businesses start low, often at zero. Assistants draw on what is searchable and well linked: directories, review platforms, "best of" articles, local news and well-structured business websites. A thin website, an incomplete Google Business Profile and a handful of reviews give an assistant little to find. Google says prominence in local ranking "is also based on info like how many websites link to your business and how many reviews you have" (Google Business Profile Help, 2026), and its AI features sit on top of the same index.
The first realistic goal is a change of state, not a number: from never being named to being named in a few answers, consistently, for the questions that matter most to you. A roofer named in three of forty answers this month and five next month is on the right track, even though the score is "only" 12.
After that, benchmark against your competitors, not an industry figure. If the three firms you lose work to are named in 40 per cent of answers and you are named in 10, that gap is your target, whatever anyone's published average says.
Be sceptical of scores quoted without a method. A "72 out of 100" means nothing unless you know how many questions, which assistants, how many runs and over what period.
How do I raise my AI visibility score?
Raise it by giving assistants more, and more consistent, evidence that you exist, what you do and that customers rate you. The score moves when the sources assistants read change.
- Make sure AI crawlers can read your site. OpenAI says "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers" (OpenAI, 2026). The AI visibility checklist has a tested robots.txt example and the other technical basics.
- Complete your Google Business Profile and gather reviews. This feeds Google's local results, AI Overviews and AI Mode, and is the quickest win for most local businesses. See Google Business Profile and reviews for AI.
- Put the facts assistants need on your own pages: one page per service, prices in plain text, service area, qualifications and an about page that proves you are real. Our guide to AI search optimisation for small businesses walks through it.
- Get onto the sources assistants already cite. Look at which sites are cited when competitors are named and you are not; a few directories, trade bodies and local publications usually account for most of them.
- Work assistant by assistant. Getting recommended by ChatGPT, how Perplexity chooses businesses and Google AI Overviews and AI Mode for small businesses cover what each one documents.
- Re-check on a schedule. Monthly at least. Nothing above shows in the score until an assistant next searches and finds the change.
To do a rough check yourself, write down 10 to 20 customer questions, ask each in ChatGPT, Perplexity and Google in a private window, note whether you were named, where, whether your site was linked and who else appeared, and divide the answers that named you by the total. Repeat monthly with the same questions. It is noisy and tedious, but free; our guide to checking whether ChatGPT recommends your business goes step by step. If you would rather a tool did the repetition, see our comparison of AI visibility tools for small businesses.
How SearchCrow helps
SearchCrow's visibility score is the share of answers that named your business, out of 100, from the questions it tracks for you across the assistants on your plan. On the free plan that is 5 questions across ChatGPT and Perplexity, checked every two weeks; paid plans track more questions, assistants and competitors (see pricing). Alongside the score, the dashboard shows share of voice against competitors, which questions name you, who is named when you are not, the sites AI cites most, and an action plan with the fixes written up. The free check runs five questions in about 30 seconds, with no card, if you want a first reading.
FAQ
What is an AI visibility score? It is the percentage of AI assistant answers, to a fixed set of questions your customers might ask, that mention your business. If you track 20 questions across 2 assistants and 10 of the 40 answers name you, your score is 25. Different tools use different question sets and assistants, so a score only means something alongside the method that produced it.
How is AI visibility measured? A tool asks a set of customer-style questions to one or more AI assistants, repeats the run on a schedule, and records for each answer whether your business was mentioned, where it appeared in the list, whether your website was cited as a source, and which competitors were named. The share of answers that mention you is the score; the rest become supporting metrics.
What is a good AI visibility score? There is no published industry benchmark, so treat any claimed one with care. Most small businesses start at or near zero because assistants lean on well-known directories, review sites and publications. The realistic first goal is to go from never being named to being named in a few answers consistently, then to track whether the trend over repeated checks is upwards.
Why does my AI visibility score change between checks? Because the answers do. AI assistants rewrite your question, search live, and generate text with some randomness, so the same question asked twice can name different businesses. Location also affects local results. A single check is a snapshot, so watch the trend over several checks rather than reacting to one movement up or down.
What is share of voice in AI search? Share of voice compares how often you are mentioned with how often each of your named competitors is mentioned, across the same questions and assistants. If the answers mention you 10 times and three competitors 15, 10 and 5 times, you hold 25 per cent of the mentions. It shows whether you are losing recommendations to specific rivals, which the headline score alone does not.
Can I check my AI visibility for free? Yes. Open ChatGPT, Perplexity or Google and ask the questions a customer would, then note whether you are named and who is. Do it monthly and keep a simple spreadsheet. Free checkers, including SearchCrow's at searchcrow.com/check, run a small set of questions for you and show a score, the assistants that named you and who was named instead.
Sources
- Google Search Central, AI features and your website
- Google Search Central, Optimizing for generative AI features on Google Search
- Google Business Profile Help, Tips to improve your local ranking on Google
- Google Search Help, Find information in faster & easier ways with AI Overviews in Google Search
- OpenAI, Overview of OpenAI crawlers
- Perplexity, Perplexity crawlers