Read the methodology
Methodology
How the AI Visibility Score is calculated: dimensions, weights and measurement rules. The formula lives in a single file (apps/web/src/lib/score.ts) and this page is generated from it.
Mention rate
Share of queries where the brand name or domain appears in the answer.
20%weight
Recommendation rate
Share of queries where the brand is explicitly suggested as an option (a mere mention does not count).
25%weight
Position
Rank among recommended businesses: 1st → 1.0, 2nd → 0.7, 3rd → 0.5, later → 0.3, absent → 0.
10%weight
Citation strength
Share of queries where the brand's domain is cited as a source.
10%weight
Entity completeness
7 Knowledge Graph signals: brand, ≥3 services, ≥1 person, ≥1 location, contact, ≥5 FAQs, ≥3 schema types.
15%weight
Local visibility
Mention rate on local-intent (district/city) queries.
10%weight
Competitive position
Your share of voice / the top share.
10%weight
Weights sum to 1.0. Total score = Σ (weight × dimension) × 100.
Per-engine score
Per engine: mention 40% · recommendation 40% · position 20%.
Measurement rules
- 01The query set is deterministic: the same brand, sector, coverage and language produce the same questions. The set is generated from the business's sector, service area and its own services, and runs up to the plan limit.
- 02Every engine's raw answer is stored as-is; never deleted or overwritten. Judging (mentioned/recommended/position) uses an LLM; without a key a rule-based match is used and the record is marked 'heuristic'.
- 03Crawled page content is untrusted input; no instruction inside it is ever executed.
- 04No data means no score; an empty state is shown. No rate is ever estimated.
- 05Finding impact points are rule-based estimates (P1: 5–7, P2: 3–4, P3: 2–3) validated through experiments.