NamedraAI Visibility Score
The AI Visibility Score is a 0-100 number that measures how often, and how prominently, AI assistants name and cite a brand when buyers ask for vendors in its category. It is computed from four observable rates across a fixed prompt set, so two months or two competitors can be compared directly.
Formula
Score = 100 × (0.40 × MentionRate + 0.25 × ShareOfVoice + 0.20 × PositionScore + 0.15 × CitationRate)- MentionRate (weight 0.40)
- Answered responses that name the brand (name, known alias or domain) divided by all answered responses. The single most important fact: were you in the answer at all.
- ShareOfVoice (weight 0.25)
- The brand's mentions divided by mentions of all tracked brands (the brand plus its competitors). Captures crowding: being named alongside six others is worth less than being one of two.
- PositionScore (weight 0.20)
- When the brand is mentioned, (K − position + 1) / K with K tracked brands, averaged over those responses. First mention scores 1.0. Never mentioned scores 0.
- CitationRate (weight 0.15)
- Answered responses that cite the brand's own domain as a source, divided by all answered responses. If no engine in a run returns citations at all, this term is dropped and the other weights are rescaled to 100.
Worked example
A compliance vendor across 450 answers: mentioned in 22%, share of voice 11%, average position when mentioned 2.4 of 4 tracked brands (position score 0.65), domain cited in 4%.
100 × (0.40 × 0.22 + 0.25 × 0.11 + 0.20 × 0.65 + 0.15 × 0.04)
= 100 × (0.088 + 0.0275 + 0.130 + 0.006) = 25How to read a score
| Score | Meaning |
|---|---|
| 0-20 | Effectively invisible. Competitors own the answers. |
| 20-40 | Named sometimes, usually late in the list, rarely cited. |
| 40-60 | A regular on the shortlist. Work is about position and citations. |
| 60-80 | A category leader in AI answers. |
| 80+ | Rare; only when a category has one dominant vendor. |
Why these weights
Mention rate carries the most weight because the commercial question is binary: on the list or not. Share of voice penalises being one of many. Position matters because buyers read the first names. Citation rate is lowest because engines often recommend a vendor while citing a review site rather than the vendor's own domain; it is a leading indicator rather than the outcome.
Limits
Answers vary between runs, so the score is computed over three runs per prompt and reported with the sample size. It depends on the prompt set; keep prompts stable month to month for comparability. It measures APIs with web search enabled, which mirror but do not replicate the consumer apps. Every component is traceable to stored raw answers.