How Pranas measures AI visibility

Pranas reports three metrics: AI Score, Coverage and Mention Rate.

Last updated: 2026-08-01

How Pranas measures AI visibility

Pranas reports three headline metrics for AI visibility: AI Score, Coverage and Mention Rate. Each answers a different question, and read together they tell you both how far your brand reaches and how strongly it lands.

The three metrics

Metric	Definition shown in the app	What it answers
AI Score	Composite 0–100 score of the quality of your presence in AI answers.	How strong are we when AI answers?
Coverage	Share of the Prompts where your brand appeared at least once.	Did we show up at all?
Mention Rate	Measure of how consistently your Brand got mentioned in AI Scans.	When answers are generated, how densely do we occupy them?

AI Score

AI Score is the headline number — a single 0–100 value that, tracked over time, tells you whether your GEO efforts are working.

For each AI Scan (one prompt run against one model, sampled across repetitions) it blends:

Rank quality (50%) — when you appear, how high in the answer you're cited, normalised against the actual length of the answer list rather than a fixed ceiling.
Appearance rate (35%) — how many of that scan's sampled repetitions mentioned you.
Rank stability (15%) — how consistent your position is across those repetitions. A brand that is reliably third is stronger than one that swings between first and tenth.

Scan-level scores are then combined weighted by appearances and scaled by Coverage, so the headline number reflects both quality and reach.

Every scan counts. A scan where your brand was never mentioned is a valid result — it is a genuine zero that pulls AI Score down, not a data point to be dropped. This is deliberate: excluding misses would inflate the score and hide exactly the prompts you most need to fix.

The app also surfaces AI Score when mentioned as a sub-line. That conditional figure tells you how good you are in the answers you do reach, which is useful for diagnosis but should never be read as the headline.

How to read your AI Score

Range	Interpretation
0–20	Largely invisible. Focus on baseline content (category pages, "what is" articles) and getting cited by influential domains.
21–40	Emerging. You appear sporadically; concentrate on the prompt types where you score lowest.
41–60	Established. You're a known option but not a default recommendation. Comparison and Problem-Solving prompts are usually the unlock.
61–80	Strong. You're in the top consideration set; protect against competitor erosion.
81–100	Dominant. Rare; usually indicates a category leader with a deep content moat.

Coverage vs Mention Rate

These two look similar and are easy to confuse, so they are worth pinning down precisely.

Coverage counts scans — one scan is one prompt run against one model. A scan either mentioned you or it didn't; there is no partial credit. Mention Rate counts answer positions — how many of the sampled slots across every scan in scope were yours. Crucially, both denominators include scans where your brand was never mentioned, so neither metric can be inflated by ignoring the misses.

Worked example. You run 10 scans, each sampling 20 answer positions (200 positions in total). Your brand appears in 4 of the 10 scans, taking 2 positions in each of those four answers.

Coverage = 4 / 10 = 40%
Mention Rate = 8 / 200 = 4%

Roughly, Mention Rate ≈ Coverage × how consistently you appear inside the answers you reach.

How to act on the pair:

Low Coverage → a reach problem. The models aren't surfacing you for those prompts at all. Fix with baseline content and citations on influential domains.
High Coverage, low Mention Rate → a consistency problem. You get named, but thinly and often just once, low in the answer. Fix with depth: stronger comparison content, more corroborating sources.
Both low → start with Coverage; density rarely improves before reach does.

What changes the numbers

Adding new prompts can move AI Score either way — new prompts you don't appear on lower it; new prompts you dominate raise it. Always interpret changes alongside your prompt library size.
Changing the date range changes all three metrics because it changes which scans are aggregated.
Country scoping changes them because different countries see different model behaviour.
Prompt-type scope matters: Unaided is the default because it's the cleanest organic signal, and Aided will almost always read higher.
Adding/removing competitors does not affect these metrics directly — they're absolute measures, not relative ones.

Common misreadings

"My AI Score dropped after a scan" — usually means a new prompt set was added or a model returned more "No brands found" responses, not that you lost ground. Compare prompt-by-prompt before reacting.
"My score is 45 but I see lots of mentions" — Mention Rate alone isn't enough; check Coverage and rank quality too.
"Score is the same but Top Brands changed" — Top Brands is relative (vs competitors), AI Score is absolute. Both can move independently.

Tracking over time

Use the AI Visibility Trends chart to see AI Score, Coverage and Mention Rate on one timeline, with per-model AI Score lines layered on top. A healthy GEO programme produces a slow, upward line. Sudden jumps usually mean a model update; sudden drops usually mean a model became more conservative or a new dominant competitor entered the set.

Share of Voice (SOV)
Top Brands widget
Dashboard filters
Prompt types and what they reveal