AI visibility: what it is and how to check your score

AI visibility is how often AI assistants such as ChatGPT, Perplexity and Gemini name your brand or link to your site when people ask the questions your buyers ask. You check it by putting a fixed set of real buyer questions to those engines and recording, for every answer, whether you were named, whether you were cited with a link, where you appeared and which competitors were named instead. An AI visibility score is a summary of those records, and it is always a sample, because the same question gets a different answer on the next run.

No score is official. Every tool's number, and your own spreadsheet, is an estimate from sampled answers, and it is worth having only if you take it the same way each month.

What is AI visibility?

AI visibility is your brand's presence inside AI-written answers, measured one question at a time. Take a prompt like "best scheduling software for a dental clinic". Each answer gives you four things to record.

Mentions and citations can split, so track them separately. An answer can cite your blog post about reducing no-shows and still recommend three competitors, because your post explained the problem while a review site's roundup supplied the shortlist. The reverse happens too. The answer names you, but the source is a Capterra category page or a Reddit thread.

A brand that gets cited but rarely named usually has useful pages and no third-party pages putting it on shortlists. A brand that gets named but never cited depends on other people's pages to speak for it. Those are different problems with different fixes, and a single blended score hides which one you have.

What does an AI visibility score measure?

Most AI visibility scores combine a few rates calculated over a prompt set, mainly how often you are named, how often you are cited and how your mentions compare with competitors'. Vendors weight these differently, so two tools give the same brand different scores.

The prompt set decides much of the score. A prompt that contains your brand name gets your brand name back almost every time, so a set padded with branded questions produces a flattering mention rate that tells you nothing. Category and problem prompts, the ones asked before a buyer has picked a vendor, are the honest test.

When anyone shows you a score, ask which prompts went in, how many runs per prompt and which engines. If they can't show you the answers behind it, the number can't tell you what to fix.

Why every AI visibility score is a sample

AI engines write each answer fresh, so one run shows what one answer said on one day. SparkToro measured how much this matters in a study run with Gumshoe, a company that sells AI tracking. About 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI results 2,961 times. Fewer than 1 in 100 runs returned the same list of brands, and the same list in the same order came up roughly once in 1,000.

The finding that rescues measurement is that how often a brand appeared across many runs held steadier than where it appeared. That shapes how you read any number:

• Treat mention rate as the headline and position as a rough guide.
• Run each prompt more than once per engine before you believe a change.
• Hold everything else still. Same wording, same engines, same settings, same logged-out state.
• Compare month against month. A drop on Tuesday is usually noise.

Answers also drift for reasons that have nothing to do with your site, such as model updates, the user's location and chat memory. More on that in why ChatGPT gives different answers to the same question.

Is there an official AI visibility score?

No. Google's AI optimization guide says no third-party tool has access to its ranking or AI systems, and calls optimizing for generative AI search "still SEO". Any score that claims to show your standing inside Google's AI is measuring sampled output, the same way your spreadsheet does.

AI search visibility in Google

Google's own number is the Generative AI performance report in Search Console, which opened to all sites on August 31, 2026. It shows impressions only, by page, country, device and date, with AI Overviews and AI Mode counted together. It tells you which of your pages appear in Google's AI features and whether that is growing. It can't tell you which questions triggered them or who appeared next to you.

Visits from ChatGPT and other assistants

For ChatGPT, watch referral traffic. ChatGPT adds utm_source=chatgpt.com to the links it cites. Perplexity and Gemini visits arrive with perplexity.ai and gemini.google.com as referrers, though desktop and mobile apps often send no referrer and the visit lands in direct. Our analytics filters for visits from AI assistants set this up in GA4 and other analytics tools.

How to check your AI visibility by hand

A manual check takes a prompt list, three or four engines, a spreadsheet and a few hours a month.

1. Write 20 to 30 prompts in your buyers' words, taken from sales calls, support tickets and site search. Most should be category or problem prompts such as "how do I cut no-shows at my practice", with a few comparisons and only a handful that name you. AI Question Finder lists the most searched questions that ChatGPT or Google AI Overviews answer about a topic, which helps fill gaps.
2. Pick the engines. Start with ChatGPT, Perplexity and Gemini, and add Google's AI Mode. Turn web search on where you can, because those answers carry sources you can record, and use a logged-out session or turn memory off. AI Overviews show up only on some searches and need their own method, covered in how to track AI Overviews.
3. Run every prompt twice per engine on the same day. Twenty prompts across three engines, twice each, is 120 answers, enough to see patterns.
4. Record each answer as one row in a sheet like the one below.
5. Total the rates per engine and overall, and list the ten domains cited most often.
6. Repeat monthly with identical prompts and settings. Add new prompts if you need to, but never reword existing ones, or the trend line breaks.

date,engine,prompt,run,named,position,brands_named,cited,cited_url,top_sources
2026-10-09,Perplexity,best scheduling software for a dental clinic,1,yes,3,"Brand A; Brand B; Acme; Brand C",no,,"capterra.com; reddit.com; brand-a.com"
2026-10-09,Perplexity,best scheduling software for a dental clinic,2,no,,"Brand A; Brand C; Brand D",no,,"capterra.com; brand-c.com"

In this made-up example, the same prompt run twice names Acme third and then drops it. That is normal, and it is why one run proves nothing. The brands_named column gives you share of voice, and top_sources tells you which pages to chase.

What AI visibility checker tools do

An AI visibility checker runs the same loop for you. It sends your prompts to the engines, uses a language model to pull brand names out of each answer, matches your domain against the cited sources and turns the counts into rates. You save hours. You also inherit choices you didn't make, so check these before trusting the number:

• Which engines it covers, and whether Google's AI Overviews and AI Mode are among them.
• How many times it runs each prompt. One run per engine is a spot check.
• Whether you can see the raw answer and sources behind every count.
• Whether you write the prompts or the tool generates them for you.
• How it reaches the engines. An API call carries no chat history or memory, so its answers compare well from run to run but may differ from what a logged-in user sees.

What to do when competitors get named instead

Start with the sources column. When the same rival shows up across engines for a prompt, open the pages those answers cite. If they are roundups, comparison articles and review listings that leave you out, getting onto those pages is a more direct fix than another post on your own blog.

If an engine cites your page but recommends someone else, your page explains the topic without making the case for your product. Add the specifics a model can repeat, such as who the product is for, what it costs and what it does that the named rivals don't. For the rest of the work on your own pages, see how to get cited by ChatGPT.

Check whether AI answers name your brand

Our AI Answer Visibility tool fills in one prompt's rows across three engines in a single run. You enter your site and one question in your customers' words, up to 300 characters. It asks ChatGPT, Perplexity and Gemini with web search on. It uses the brand name you type, or reads it from your home page's markup and falls back to your domain.

For each engine, the report shows whether the answer names your brand and in what place, such as 3rd of 8. It also shows whether the answer cites your site, which of your URLs it links and the full answer text. Separate tables list up to 10 brands and 10 sites the answers name and cite instead. A run costs 120 credits.

Each run takes one answer per engine, so treat it as one sample and rerun the prompts that matter most. It doesn't cover Google AI Overviews or AI Mode, and it doesn't track answers over time. For Google, use the Search Console report.