The AI Visibility score | Recited AI Docs

The formula is public

Your AI Visibility score is a single 0 to 100 number summarizing how present your brand is in AI answers. The formula is public, because a score nobody can explain does not survive its first executive review:

AI Visibility = 55% topic coverage + 45% mention strength vs the median competitor

Both parts are shown next to the score on the Overview stat strip, so you can always see which one moved.

Topic coverage (55%)

Topic coverage asks: across the prompts you track, in how many do you appear at all?

It is weighted heaviest because absence is the biggest problem in AI visibility. A brand mentioned somewhere in most answers is in a fundamentally better position than a brand that dominates three prompts and is invisible everywhere else. Coverage grows when you become part of answers you were missing from, which is exactly what opportunities and missing citations target.

Mention strength (45%)

Mention strength asks: when you do appear, how strong is your presence compared to the median competitor?

Measuring against the median matters. Comparing against the single best competitor would make the score swing on one rival's good week; comparing against the average would let a long tail of tiny brands flatter you. The median competitor is the honest midpoint of your actual market.

If your mentions are stronger than the typical tracked competitor's, this component is high. If you appear but always briefly and last, it is low, even with decent coverage.

Reading the score

Score up, coverage up: you entered answers you were missing from. The healthiest kind of gain.
Score up, strength up: answers now feature you more prominently where you already appeared.
Score down, nothing you did changed: often a competitor got stronger, since strength is relative, or answers shifted to sources that do not know you. Check the leaderboard and momentum.

What the score is not

It is not share of voice, which is a pure mention-count ratio; see share of voice. It says nothing about tone; that is sentiment. And like every metric in Recited, it is computed from stored answers you can open and read, so when the number surprises you, the answer log has the receipts.