Methodology

A weighted 0–100 measure combining mention frequency (30%), recommendation frequency (30%), answer position (15%), citation frequency (15%) and question value weighting (10%). It is our composite, not a metric published by any AI provider.

Your brand mentions divided by all brand mentions detected across the same set of answer runs. Brands outside your tracked competitor set are aggregated into 'Others'.

The share of answer runs that cite at least one URL you own, or a third-party URL whose cited passage references your brand.

Questions are derived from your products, services, categories, customer problems and competitor set, then classified by intent. You can edit, delete, import and regenerate them — the question bank is yours, not a fixed list.

Each run stores the full answer, detected brand mentions with ordinal position, competitor mentions, cited URLs and a sentiment classification. Nothing is inferred without a stored, timestamped run behind it.

Recommendations are produced only when a measured signal gap exists — a missing page for a query pattern, a source that cites competitors and never you, an entity association absent from answers, or a crawl-detected technical fault. Each carries the evidence that triggered it.