A shared vocabulary for measuring, learning and improving.
The observed presence of an entity or its content in AI-mediated answers within a defined context.
A saved answer names your product. That is one observation for one prompt, product, language, and time; it does not describe all answers.How a page meets explicit editorial or technical criteria before separately observing its external visibility.
Suprara checks public-page basics with recorded evidence and coverage. A strong result does not show whether any AI answer cited the page.An identifiable person, organization, product, place, or other subject.
A name plus a category, official website, and service area helps distinguish a fictional Luma salon tool from unrelated businesses called Luma.An appearance of an identifiable entity in an answer.
A comparison names your product among several options. Count a mention, then separately classify whether it is neutral, favorable, or critical.A visible reference or link to a source in an answer.
An answer links your installation guide without naming your brand in its prose. That can be a citation without a brand mention.Finding candidate information that may help answer a question.
A question about software costs may lead a system to retrieve pricing and plan documentation. Retrieval alone does not prove a page was cited.Connecting an answer to external information used as supporting context.
An explanation can use a current policy page as context. The resulting claims still need checking against the policy’s scope and exceptions.Issuing related searches across subtopics to develop an answer.
A broad comparison may involve questions about price, compatibility, and support. Google documents this technique for its AI search features.Whether a crawler can access the intended public content under the relevant technical controls.
A page works for a signed-in editor but may not be available to a crawler. Ask a technical owner to verify access rather than assuming it.A search system’s processing and inclusion of content in its searchable index.
A public URL is not proof of indexing. Keep “published,” “accessible,” and “indexed” as distinct states in a technical review.Machine-readable information using a defined vocabulary or format.
Organization markup can describe a real organization. Its details should agree with the visible page rather than add imaginary qualifications.Original documentation, data, or testimony directly relevant to a claim.
Use official product documentation to confirm a supported feature. That same source does not independently establish market-wide superiority.The total eligible observations used to calculate a rate.
Three mentions across ten completed test answers gives a 30% rate for that sample. Report failed requests and exclusions separately.A fixed scoring method that returns the same result for the same inputs and version.
Identical captured HTTP/HTML evidence and rubric versions yield identical Suprara scores. Scanning the same URL later can produce different evidence, so preserve both scan records.Information supplied by the user that has not necessarily been independently verified.
Saying “our page is indexed” is a declaration until supported by appropriate evidence. Suprara fetches observable public-page signals, but a successful fetch alone does not verify indexing.A relationship in which one change produces another, requiring stronger evidence than coincidence.
More mentions after a rewrite do not alone prove the rewrite caused them. Provider changes or different test conditions may also explain the result.