What Is AI Visibility? Metrics and How to Measure It

What is AI visibility

AI visibility is how often, how prominently, and how favorably a brand appears when people ask AI engines like ChatGPT, Claude, Gemini, Perplexity, and Grok for a recommendation. It is a composite of five measurable metrics: mention rate, position in the answer, share of voice, sentiment, and citation presence. A brand with high AI visibility shows up often, near the top of the answer, described well, alongside a source link, across more than one engine.

Mention rate

Mention rate is the percentage of AI answers, across a set of buyer-intent questions and engines, in which your brand is named at all. It is calculated as the number of answers that mention the brand divided by the total number of answers checked.

Worked example: our sample report asked five engines ten buyer-intent questions about project management software, fifty answers in total. Notion was named in 19 of the 35 category answers, for an overall mention rate of 54 percent. Broken out by engine, the rate ranged from 43 percent on Gemini, Perplexity, and Grok, to 57 percent on ChatGPT, to 86 percent on Claude. That per-engine spread is normal, which is why a mention rate is only meaningful when it states how many engines and how many prompts it is averaged across.

Position in the answer

Position is the ordinal place a brand occupies among the brands named in a single AI answer: the first brand mentioned is position 1, the second is position 2, and so on. An average position is the mean of those ordinals across every answer that mentioned the brand.

Worked example: if an engine lists "Asana, Notion, and ClickUp" as options, Notion's position in that answer is 2. If Notion appears in five answers at positions 1, 2, 2, 4, and 3, its average position across those five is 2.4. Lower is better, and position only exists for answers where the brand was mentioned at all, so it is always read alongside mention rate, not instead of it.

Share of voice

Share of voice is the percentage of all brand mentions, across every competitor named in a set of answers, that belong to your brand. It is calculated as your brand's mention count divided by the combined mention count of every brand named, including you.

Worked example: across the fifty answers in our sample report, the engines most often recommended Asana, Monday.com, and ClickUp alongside Notion. If those four brands together were named 80 times and Notion accounted for 27 of those mentions, Notion's share of voice would be about 34 percent, a figure that says more about competitive standing than mention rate alone, since it accounts for how crowded the answer is.

Sentiment

Sentiment is whether an AI engine describes your brand positively, neutrally, or negatively at the moment it mentions you. It is scored per mention, then rolled up as a percentage positive, neutral, and negative across all mentions.

Worked example: "Notion is the most flexible option for teams that want an all-in-one workspace" scores positive. "Notion, though it can be overwhelming for people who just want simple task tracking" scores negative. A brand can have a high mention rate and still lose deals if the sentiment attached to those mentions skews negative.

Citation and source presence

Citation presence is whether an AI engine names or links to your own domain as the source behind a claim, as distinct from simply naming your brand in prose. An engine can mention "Notion" ten times without ever citing notion.com, citing a review site or a comparison blog instead every time.

This matters because the domains an engine cites for a category are usually a small, repeatable list, third-party review sites, comparison roundups, and a handful of category leaders' own pages, and being outside that list means your brand's visibility depends entirely on what other people have written about you. A full report lists exactly which domains an engine cited for your category, under "where AI looks," alongside the mention data.

Why AI visibility is not a search ranking

Search rankings are stable enough to track daily because Google returns a fixed list of ten results for a fixed query. AI visibility has no fixed result set: an engine can name two competitors in one answer and five in the next, for the exact same question. Answers also vary run to run on the same engine, because generative models sample from a distribution of likely next words rather than looking up a static list. And engines disagree with each other: our own data shows the same brand, same category, same day, mentioned in 43 percent of Grok's answers and 86 percent of Claude's. None of that is a bug. It just means a single check, on a single engine, on a single day, is not a measurement.

How to measure AI visibility

Measuring AI visibility means asking the same set of buyer-intent questions across multiple engines, more than once, and scoring every answer for the five metrics above. A useful measurement states its inputs: how many prompts, how many engines, how many runs, so the resulting numbers can be trusted and compared over time.

You can see the full method applied to a real brand in our sample report, or get one for your own brand: a free scan checks one prompt across two engines with no email required, and the $19 report runs ten buyer-intent prompts across all five engines, fifty answers, scored for every metric on this page, plus a competitor leaderboard and specific recommendations. Full scoring definitions are documented at methodology, and the companion guide on generative engine optimization covers how to actually improve these numbers.