Last updated 2026-06-30
MethodologyHow we measure AI visibility
CiteMatrix.ai scores are computed from real responses collected from leading AI models — never estimated, modeled or invented. This page documents exactly how the AI Visibility Score and AI Share of Voice are calculated, so you can trust (and verify) every number we publish.
1. Data collection
We maintain a library of real buyer-intent prompts for each category — the kinds of questions B2B buyers actually ask, like "what's the best CRM for a startup?" We run these prompts across ChatGPT, Claude, Gemini and Perplexity on a recurring basis and capture the full responses, including which companies are mentioned and in what order.
2. The AI Visibility Score (0–100)
Each company's score is a weighted blend of five components, each computed from the collected responses:
| Component | Weight | What it captures |
|---|---|---|
| Mention frequency | 35% | How often a company is named across the tracked prompt set, relative to its category. |
| Average position | 25% | Where the company appears in AI answers — first recommendation, third, or buried. |
| Model coverage | 20% | How many of the major models (ChatGPT, Claude, Gemini, Perplexity) mention it. |
| Category relevance | 10% | Whether mentions occur for the prompts that actually match the company's category. |
| Sentiment | 10% | How positively the company is described — recommended, mentioned neutrally, or cautioned against. |
3. Position scoring
Position matters: being the first tool an AI names is worth far more than being mentioned last. We convert each appearance into a position score:
| Appearance | Position score |
|---|---|
| 1st recommendation | 100 |
| 2nd | 85 |
| 3rd | 70 |
| 4th | 55 |
| 5th | 40 |
| Mentioned, not ranked | 25 |
| Not mentioned | 0 |
4. AI Share of Voice
AI Share of Voice measures how much of a category's AI attention a single company captures. It's the company's weighted mentions divided by the total weighted mentions for all companies in that category. A share of voice of 20% means one in five weighted AI mentions in that category point to this company — a zero-sum view of the category: as one brand gains, others lose.
5. What we will never do
- We never invent or estimate a score to fill a gap.
- We never show rankings for a category before we've collected real responses.
- We never publish pricing we can't verify from the vendor.
- Company metadata (HQ, funding, ICP) is researched and real, not generated.
Run a free snapshot computed from real model responses, or book a full audit to see exactly where you stand against competitors.
Run a free AI-visibility auditFAQ
- Does CiteMatrix.ai ever estimate or model scores?
- No. If we haven't collected real model responses for a company or category, we show a coming-soon state — never a placeholder or estimated score. Every number reflects real data.
- Which AI models does CiteMatrix.ai measure?
- ChatGPT, Claude, Gemini and Perplexity — the assistants B2B buyers most often ask for recommendations. Model coverage is one of the five scoring components.
- How is AI Share of Voice different from the AI Visibility Score?
- The AI Visibility Score (0–100) measures one company's standing. AI Share of Voice measures how much of a category's total weighted AI mentions that company captures — a zero-sum view across the category.