Why it's a share across a panel, not a single query
One question asked once is close to useless. AI answers are stochastic — the same prompt returns you in one draw and omits you in the next, and different engines disagree with each other. Reading a single answer tells you nothing durable.
A share fixes that. Freeze a question set, run it through every engine in the panel, sample each query k times, and count how often your brand appears out of the total answers produced. The rate that comes out is stable: it moves when your standing actually changes, not when a model happened to roll differently that afternoon. That is the difference between a metric you can trend and a vanity 'we appear sometimes.'
Instrument: a frozen five-engine panel — ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews — each query sampled k times. Freezing the panel and the question set is what makes two scans comparable.
Worked example — our own day-0 index
Our first self-scan (2026-07-19, frozen five-engine panel) asked the vendor-selection question set and counted the share of answers that named each tool. This is citation share read straight off the panel — three incumbents holding real share, and the AI Citation Institute at zero on day zero. The named tools are not our clients; this is just what the panel returned.
| Tool | Citation share |
|---|---|
| Semrush | 59.4% |
| Profound | 48.4% |
| Otterly.AI | 43.8% |
| The AI Citation Institute | 0.0% |
Share = fraction of panel answers naming the tool on the vendor-selection question set. Named tools are not The AI Citation Institute clients.
Mention share vs active-recommendation share
One citation-share number hides a split worth pulling apart: getting named is not the same as getting recommended. Mention share counts every answer where the brand shows up at all. Active-recommendation share counts only the answers where the engine actually puts the brand forward as the answer.
Gavelist, one ongoing client, makes the gap concrete: it reached a 58.5% mention rate against the nearest rival's 38.8% — a clear lead on being named. But active recommendation, being put forward rather than merely mentioned, sits around 13%. That gap between mention and recommendation is the frontier, and it's where the work is.
Gavelist figures are for one ongoing client and illustrate the mention-vs-recommendation split, not a benchmark for every account.
Common questions
How is AI citation share different from SEO rank?+
Rank measures where a page sits in a list of links. Citation share measures how often AI answers name or cite your brand across a fixed question set and engine panel. There's no ordered list — either the answer names you or it doesn't, and the metric is the rate across many answers.
Why sample each query several times instead of once?+
AI answers are stochastic: the same prompt names you in one draw and omits you in the next. Sampling each query k times and averaging turns that noise into a stable rate, so a change in the number reflects a real change in standing rather than a lucky or unlucky roll.
Is mention share the same as recommendation share?+
No. Mention share counts every answer where the brand is named at all. Recommendation share counts only answers where the engine puts the brand forward as the answer. The two rates often differ by a lot — Gavelist reached 58.5% mention but about 13% active recommendation — and the gap is usually the real work.
keep reading
See what AI says about your brand right now — free, on a live engine.
Scan your domain