Intro
“Are we visible in AI?” is a binary question. Useful answers are not.
The Capston AI visibility scoring system replaces vague impressions with a structured measurement: which prompts surface the brand, which competitors are named first, which sources are cited, which facts are wrong, and where commercial risk concentrates.
This page explains what the score covers, how it is built, and what it tells a brand to do next.
What the score covers
Eight dimensions, scored independently and then aggregated.
- Brand presence — does the brand appear in the answer at all?
- Answer position — first named, second, third, or buried.
- Citation share — proportion of prompts where the brand’s own domain is cited.
- Source quality — trust level of the URLs AI engines reuse.
- Competitor dominance — which competitors win, and on which prompts.
- Fact accuracy — claims about the brand: correct, wrong, missing.
- Sentiment — neutral, positive, negative descriptors.
- Commercial risk — OTA capture, aggregator routing, intermediary substitution.
Each dimension is measured against the same prompt set, the same competitor set, and the same engine set. That is what makes the score comparable over time and across brands.
How the score is built
The prompt set is the foundation.
For each brand, Capston Core co-designs a prompt library across four intent buckets:
- Discovery — broad queries that surface a shortlist
- Comparison — head-to-head and “best of” queries
- Trust — review-driven, reputation-driven, accuracy-driven queries
- Conversion — branded queries where the answer drives or breaks a booking
Each prompt is run across the relevant AI engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, and others depending on the market). Each answer is captured, dated, and stored with model version metadata.
From those captures, the scoring engine produces eight dimension scores and one composite. The composite is not the headline — the eight dimensions are. The composite is a navigation aid.
What the score tells you
A score is only useful if it leads to a decision.
Each Capston Core score comes with:
- The biggest gap (which dimension is weakest, on which prompts)
- The easiest gain (which dimension can move first, with the least effort)
- The biggest risk (which dimension exposes the brand commercially today)
- A recommended next move
This is what differentiates scoring from reporting. A report shows the state. A score sequences the work.
How this fits into Capston Core
Scoring is one of the six things Capston Core owns. The score relies on the AI answer evidence layer for traceability, on the Capston Core methodology for the five-stage process, and on the Capston QA standards for consistency across clients and partners.
→ Back to Capston Core
FAQ
How many prompts go into a score?
Typically 40 to 80 prompts, depending on portfolio size and market scope. Locked once agreed.
Which engines are scored?
The engines that matter for the brand’s markets. Default set: ChatGPT, Perplexity, Google AI Overviews, Gemini. Adjusted per market.
How often does the score change?
Models update, content moves, competitors react. Quarterly retests are the standard cadence; high-stakes accounts move to monthly.
Can the score be compared across brands?
Within the same vertical and prompt set, yes. Cross-vertical comparison is misleading because the buying journey differs.
Final CTA block
Score how AI engines describe your brand.
Score your brand
Read the methodology