How BrandCitation Scores AI Visibility
Inside the BrandCitation scoring model: prompt design, mention detection, citation rates, share-of-voice, and the 0–100 visibility score.
Every BrandCitation report includes an AI visibility score from 0–100. This post explains how that number is calculated, what each metric means, and how to interpret results.
For the full technical methodology, see /methodology.
The core idea
We send buyer-intent prompts to AI engines (Gemini, Perplexity, OpenAI-powered models) and analyze whether your brand appears in responses. The score aggregates mention rate, citations, competitor presence, and technical readiness.
Prompt design
Prompts are grouped by intent:
- Brand — "What is {brand}?", "Is {brand} legit?"
- Category — "Best {category} in {market}"
- Competitor — "Alternatives to {competitor}"
- Problem/solution — "I need a tool for {job-to-be-done}"
- Local/MENA — Arabic and English regional variations
The free scan runs 3 prompts on Gemini. The paid audit runs 15–50 prompts across all engines.
Mention detection
We detect brand presence via:
- Exact brand name match in response text
- Case-insensitive match
- Domain string match (your website URL in the answer)
A brand mention rate = mentions / total prompt-engine checks.
Citation detection
Citations are counted when:
- Perplexity (or similar) returns your URL as a source
- Response text includes your domain
- A cited page's domain matches yours
Citation rate = citations / total checks. Often lower than mention rate — many AI answers name brands without linking sources.
Share of voice
For each prompt set:
Your mentions / (Your mentions + All competitor mentions)
If you are mentioned 2 times and competitors appear 13 times total, share-of-voice = 13.3%.
Sentiment and position
We classify sentiment as positive, neutral, or negative based on how the AI frames your brand. Position tracks list order when the AI returns ranked recommendations (1st vs. 5th vs. absent).
Technical GEO score (separate)
The GEO checklist scores technical readiness 0–100:
/llms.txt— 25 points- AI bots allowed — 15 points
- sitemap, schema, meta, canonical, Open Graph — remaining points
This is reported alongside prompt visibility in full audits.
Score bands
| Score | Label | Meaning |
|---|---|---|
| 0–25 | Critical | Rarely or never mentioned |
| 26–50 | Weak | Occasional mentions, competitors dominate |
| 51–75 | Moderate | Present but inconsistent |
| 76–100 | Strong | Regular mentions and citations |
See a real example in our sample report.
Limitations
API results may differ from consumer ChatGPT or Perplexity UI. We use APIs for scalable, repeatable trend tracking and state this clearly in every report.
Try it
- Free scan — 3 prompts, snapshot score
- Full audit — complete scoring + 30-day action plan
FAQ
What is a good AI visibility score? Above 75 is strong, 51–75 is moderate, 26–50 is weak, and below 25 is critical. But the score matters relative to your competitors — a 40 in a category where everyone scores 25 is actually leading.
How is the score calculated? BrandCitation aggregates mention rate, citation rate, share of voice, sentiment, position, and language coverage across a fixed prompt set, then maps the result to a 0–100 scale. Technical GEO readiness is scored separately and shown alongside.
How many prompts does BrandCitation test? The free scan runs 3 prompts on Gemini. The full audit runs 15–50 prompts across Gemini, Perplexity, and OpenAI-powered models, in English and Arabic where relevant.
Why do API results differ from the ChatGPT UI? APIs return cleaner, more deterministic outputs that scale for trend tracking. The ChatGPT and Perplexity UIs add personalization and vary session-to-session. BrandCitation uses APIs for repeatable measurement and discloses this in every report.
How often should I re-score? Monthly for trend tracking. Weekly if AI discovery is an active lead channel. Use the same prompt set every time so the trend is comparable.