How we measure

Metrics

Mentions

Successful answers in which the brand name, an alias or the domain appears, divided by all successful answers.

Recommendations

Successful answers that present the brand as an option (in a list, highlighted or with recommendation wording), divided by all successful answers. A person reviews the automatic interpretation.

Domain citations

Successful answers in which the company’s domain is among the cited sources, divided by all successful answers. Sources consulted by search but not cited are not counted.

Inaccuracies

Reviewed answers with at least one verifiable error about the company, divided by reviewed answers that mention the company.

Technical gaps

Separate findings with evidence: HTTP status, robots.txt, noindex, canonical, sitemap, titles, structured data and contact details. Technical indexability is not confirmed indexing.

What we store for every observation

The exact question, platform, model and version information, date, language, requested market and whether the market setting was applied, search mode, the answer, source URLs, execution status, duration and cost.

Errors and sample size

Failed or skipped requests are never counted as “not mentioned”. We show the sample size and the variation between runs. A handful of runs does not give statistical precision, and we do not claim it does.

An API measurement is not a copy of the app

API answers are labelled as API measurements. They are not an exact reproduction of the consumer ChatGPT experience, and the Gemini API is not Google AI Overviews.

Comparing over time

Question sets are versioned. Before/after comparisons use the same settings and flag any change in model or methodology.

What we do not do

We do not promise positions in AI answers, offer fake reviews, link exchanges or spam content, or change client websites automatically.