Glossary
What each term means, in plain words. Where a definition depends on a setting, the current value is shown with its source.
- Mention rate
- The share of valid AI answers that name your brand. Each answer counts once, however many times the brand appears in it.
- Top-pick rate
- The share of valid answers that present your brand as the first or best choice.
- Confidence interval
- The range the true rate most likely sits in, given how many answers we collected. More answers give a narrower range.
- Confidence level of every interval (%): 95
- Verified
- A number that passed every check of the Verification Gate: enough answers, a narrow enough interval, no engine canary failures, no unexplained model change, and agreement with an independent second calculation.
- Minimum valid runs before any rate is shown: 5 · Maximum 95% interval half-width (percentage points) for Verified: 7.5
- Provisional
- A real measurement that did not meet every Verified check, usually because the interval is still wide. Useful for direction, not for decisions on its own.
- Insufficient data
- Too few valid answers to show a rate at all. We show the count instead of a number.
- Minimum valid runs before any rate is shown: 5
- Verification Gate
- The checks every check passes through before its numbers are published: quarantine of failing engines, drift detection, a second independent calculation, and statistical thresholds.
- Sealed snapshot
- An immutable record of one check: the exact runs used, the code version and a content hash, so anyone can re-verify the numbers later.
- Run
- One question asked to one AI engine once, stored with the exact prompt, the full answer, the model version and the cost.
- Model drift
- An engine changed its model or behaviour. Movement on that engine is flagged as drift, not as a change in your visibility.
- Canary question
- A question with a known answer asked alongside yours. If an engine gets canaries wrong, its answers are quarantined for that window.
- Claim accuracy
- Of the checkable things engines say about you (price, founding year, features), the share that match the facts you recorded.
- AI Visibility Score
- A summary score out of one hundred, built from five measured shares: mentioned, recommended, share of voice, not negative, and accurate claims. It comes with a range and a quality badge, and is withheld when any part lacks data.
- Visibility Score weight: mentioned in answers (points out of 100): 40 · Visibility Score weight: recommended when mentioned (points out of 100): 20 · Visibility Score weight: share of voice among tracked brands (points out of 100): 20 · Visibility Score weight: mentions that are not negative (points out of 100): 10 · Visibility Score weight: checkable claims that are accurate (points out of 100): 10
- Free Scan
- A quick public check with a small sample. Its results are always provisional.
- Questions asked in a Free Scan: 5 · Times each Free Scan question is asked per engine: 3 · Engines asked in a Free Scan: 2
- Gap
- A question where you are named in fewer than half of the answers, often while a competitor is named instead.
- Roadmap
- A plan of initiatives built only from verified findings, each linked to the answers behind it and re-measured after you finish it.
- Closed-loop check
- A re-measurement of the same questions after an initiative is done. The before/after change is reported as an association, not proof of cause.
- Days after an initiative is marked done before its after-measurement check runs: 14
- Live Example
- A real brand that agreed to have its results shown publicly, so you can see the product with real data before adding your own.