GrowthGPTGrowthGPT
Start Building
Methodology

How the Score Is Calculated

The full scoring model behind the 0 to 100 AI Visibility Score: five weighted signals, sample-level averaging, and the tier scale.

Last updated July 8, 2026

The AI Visibility Score is a weighted average over real AI answers, not a vibe. This page documents the exact model: the scoring unit, the five signals and their weights, the aggregation formula, and the tier scale. The same formula runs for every brand, every scan.

The scoring unit: one sample per query and engine

Every combination of one query and one engine is a sample. With 25 queries and 4 engines, a full scan has up to 100 samples. Scoring per sample instead of per query is deliberate: it rewards cross-engine consistency. A brand that all four engines recommend outscores a brand that only one engine knows about, even if both "appeared" for the same number of queries.

A sample where your brand is not mentioned scores zero. A sample where it is mentioned scores the weighted sum of five signals.

The five signals and their weights

SignalWeightHow it is graded
Mention40%Does your brand appear in the answer at all? Binary: 1.0 if named, 0 if not.
Position20%First recommendation 1.0, top three 0.8, part of a longer list 0.5, mentioned but buried 0.2.
Citation strength20%The primary recommendation 1.0, one of several options 0.6, a passing mention 0.3.
Sentiment10%Positive framing 1.0, neutral 0.5, negative 0.
Source authority10%Whether your own domain appears in the top live web sources retrieved for that query. Requires a website URL in the scan.

Mention carries the largest weight because in AI answers it is the binary that matters most: there is no page two, so being named at all is the first battle. Position and citation strength together carry 40% because a buried, passing mention is worth far less than being the engine's primary recommendation. Sentiment keeps a negative mention from counting as a win. Source authority ties the score to something you directly control: your own domain's presence in the web sources AI engines read.

From samples to the 0 to 100 score

The aggregation is a plain average, published in full:

sample_score = 0   if brand not mentioned in the answer

sample_score = 0.40 * mention          (1.0)
             + 0.20 * position_score   (1.0 / 0.8 / 0.5 / 0.2)
             + 0.20 * citation_score   (1.0 / 0.6 / 0.3)
             + 0.10 * sentiment_score  (1.0 / 0.5 / 0.0)
             + 0.10 * authority        (1 if your domain is in the
                                        query's top web sources, else 0)

overall = average(all sample_scores) * 100

A worked example: suppose your brand appears in an answer as one of several options (citation 0.6), listed in the top three (position 0.8), framed neutrally (sentiment 0.5), and your domain was in the retrieved sources (authority 1). That sample scores 0.40 + 0.16 + 0.12 + 0.05 + 0.10 = 0.83. If your other 99 samples averaged the same, your overall score would be 83.

Because unmentioned samples score zero and stay in the denominator, scores are honest about absence. A brand mentioned brilliantly in 10 of 100 answers still scores low, because 90 zeros drag the average down. That is by design: the score measures visibility, not peak performance.

The tier scale

TierScore rangeWhat it means
Invisible0 to 15AI engines rarely or never mention you.
Emerging16 to 35You appear occasionally, but inconsistently.
Visible36 to 55You show up regularly, not dominantly.
Prominent56 to 75AI frequently recommends your brand.
Leader76 to 100A top recommendation across most queries.

Most established brands land in Emerging or Visible on their first scan. Leader scores are rare and typically belong to category-defining brands, because they require being recommended consistently by all four engines across all five query intents.

Supporting metrics in the breakdown

  • Mention rate is query-level, not sample-level: the share of the 25 queries where your brand appeared in at least one engine's answer. This is the "appeared in X of 25 queries" number in your report, and it will usually be higher than the overall score.
  • Position, citation, and sentiment averages are computed over mentioned samples only, so they describe the quality of the mentions you do have rather than repeating the absence penalty.
  • Authority score is the share of queries where your own domain appeared in the top live web sources.
  • Share of voice is your mentions divided by all brand mentions across every answer in the scan. The same tally builds your competitor table. See Reading your report.

See your own AI Visibility Score

Free scan across 4 AI engines and 25 buyer queries. No signup required.

Check your score