Metric definitions

Exactly how named, cited, recommended, position, share of voice and the 95% interval are calculated — and where they can be wrong.


Every metric in Citeroot has a stated numerator and denominator. This page is the reference. All of them are computed from completed answers; answers that failed are not counted as absences, and they don't use your allowance.

#Named (mention rate)

Named = completed answers that name the brand ÷ completed answers

An answer names your brand when its text contains your brand name, one of the other names you added, or your domain, as a whole word (so "Fern" does not match inside "Fernhill"). Matching ignores case. Citation markup such as [1] or ([site](url)) is removed first, so a brand that only appears inside a link label isn't counted. Matching looks for whole words, so it works best for languages that put spaces between words; in Chinese, Japanese or Thai a brand written directly against other characters may be missed — read those answers rather than relying on the rate.

#Your site cited

Cited = completed answers that cite a page on your domain (or a subdomain) ÷ completed answers

Only sources the engine actually cited count. Pages it merely looked at ("consulted") are stored but not counted. A competitor's page about you is not an owned citation.

Recommended = completed answers that recommend the brand ÷ completed answers

This is a heuristic. An answer counts when a sentence naming your brand also contains a recommending phrase ("I recommend", "best choice", "well suited", "top pick", "stands out" …) and no negation ("don't recommend", "avoid", "not ideal"). The sentence is stored as the evidence and shown on the answer page. It will sometimes be wrong; read the answer before you act on it.

#Position

Position is the order in which brands are first named, among the brands you track (you plus your competitors): 1 means you were named before any tracked competitor. It is reported as an average next to the rates, not blended into them. Brands you don't track aren't counted.

#Share of voice

Share of voice = answers naming you ÷ (answers naming you + Σ answers naming each tracked competitor)

It's an answer-level share: each answer counts once per brand, however many times it repeats the name. It needs competitors to be meaningful — add them in Project settings. Adding or removing a competitor changes everyone's share.

#95% intervals

Rates are shown with a Wilson 95% interval, which widens when there are few answers. "Named 40%, interval 25–57%" means the data fits anywhere in that range. Intervals describe sampling noise in the answers we collected; they don't capture bias from your choice of questions or engines.

#What isn't a metric

Citeroot doesn't produce a single composite "AI visibility score". A composite hides which signal moved. Use the separate measures, and read the answers.

#Not built yet

Accuracy of what the answer says about you, and excluding individual runs from calculations, are not available.

Last updated Oct 8, 2026 · Suggest an edit