AI search visibility metrics: the KPIs that matter, and how to calculate them

Which AI search visibility metrics matter?

The AI search visibility metrics that matter are the ones that show whether AI assistants name you, how they describe you, and what that leads to. In practice that is five KPIs: mention rate, share of voice, citation rate, gap rate and description accuracy, plus outcome metrics such as calls, bookings and AI referral visits. Track them weekly against a fixed set of prompts.

AI search visibility is different from search rankings because an answer usually names two to five options in a paragraph. There is no page two. You are either in the answer or you are not, so the first metric is simply how often you are in it.

The core AI visibility KPIs, defined

Each KPI below is calculated over one prompt set, one engine (or all engines combined) and one period. Keep those three fixed when you compare weeks.

Mention rate        = answers that name you / total answers
Share of voice (AI) = your mentions / mentions of all tracked brands
Citation rate       = answers citing your domain / total answers
Gap rate            = prompts where a competitor is named and you are not / total prompts
First-mention rate  = answers where you are named first / answers that name you
Description accuracy = mentions with correct facts / answers that name you

Total answers = prompts x engines x runs (x locations, if you track per location)

The share of voice (AI) glossary entry explains that metric in more depth, and AI visibility defines the umbrella term.

Quality metrics: position, sentiment and sources

Being named is the start. How you are named matters too.

• First-mention rate. Of the answers that name you, how often you come first. Many readers act on the first option.
• Sentiment and framing. Whether the answer describes you neutrally, positively or with a caveat ("some reviews mention long waits"). Record the caveat word for word.
• Source coverage. Of the sources the engines cite for your prompts, how many include you with correct facts. This is the metric most directly under your control.
• Engine coverage. How many of the engines you track name you for a given prompt. Being named in one engine and missing in four is a source problem, not a brand problem.

Outcome metrics: what AI visibility leads to

Visibility KPIs need a link to business results, or nobody will fund the work. Most AI answers send no click, so look beyond referral traffic.

Two of those sources are new and free. Bing's AI Performance report, in public preview since February 2026, shows total citations, average cited pages and sample grounding queries for Copilot and AI summaries in Bing (Bing Webmaster blog). Google says AI Overviews and AI Mode traffic is included in Search Console under the "Web" search type (Google Search Central).

For locations, tie calls back with call tracking: a tracked number per location shows whether demand moved after visibility did.

A weekly AI brand monitoring dashboard

Put everything on one page. One row per location (or one row for the brand), one column per KPI, four-week averages, and a short list of the top gaps with competitor names.

Invented example figures to show the layout. Your numbers depend on your prompt set, engines and competitors.

Which AI visibility KPIs to report to whom

Different people need different numbers. Sending everyone the full dashboard means nobody reads it. Match the KPI to the decision each person makes.

Keep the executive view to one chart and one sentence: "Share of voice for our top 20 questions is up from 18% to 24% over the quarter; calls from the three weakest locations are flat, and here is the plan." (Invented figures, to show the format.) Put the detail in the appendix.

How to keep AI visibility metrics honest

AI answers are not deterministic. The same prompt can give a different answer an hour later, from a different city, or for a different account. Your metrics are samples, so treat them like samples.

• Keep the prompt set stable. Change it once a quarter, and mark the date on your chart.
• Run each prompt more than once. Several runs per prompt per week give steadier numbers than one.
• Fix the location. Run local prompts from, or set to, the location you are measuring.
• Separate branded and unbranded prompts. Branded prompts ("is X good") inflate mention rate. Report them separately.
• Compare engines separately before combining. A combined score can hide that one engine never names you.

Good metrics start with good prompts. See how to choose prompts to track, then compare trackers in our best AI visibility tools roundup.

Start measuring this week

You can calculate every KPI on this page by hand for one location and 20 prompts. Run the free checker to get your first data points, then set up the sheet. When the sheet gets too big, a tool takes over the running and scoring. For what to do with the gaps, read how to improve brand visibility in AI answer engines and the generative engine optimization guide.

Every KPI on this page, per location, every week

Visibility by Locus Intelligence runs your prompts in ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI Overviews, calculates mention rate, share of voice and gaps per location, and shows the sources behind each one.

Book a demo

AI visibility checker

Business and city, or a websiteCheck

Free, no account. You get whether ChatGPT, Gemini, Perplexity and Claude name you, how they describe you, and who they pick instead.