Entity SEO + Wikidata matching

What the Entity SEO audit does

/dashboard/sites/<id>/audit/entities extracts named entities from every page, classifies them, and matches them against Wikidata's knowledge graph.

Extraction modes

• Local (all tiers) — regex-based NER for capitalised multi-word names with particle-tolerant pattern (handles "de la", "van der", "von" prefixes), classified into Person / Organization / Place / Product / Brand / Event / WorkOfArt / Other via ORG_SUFFIXES, PERSON_TITLES, and PLACE_HINTS dictionaries
• AI-enhanced (Agency+) — Sonnet 4.6 strict-JSON extraction with top-40 cap and ≥2-occurrence floor

Wikidata matching

For each entity with ≥3 occurrences, we run a SPARQL query against Wikidata's public endpoint with an exact/substring/fuzzy confidence ladder (0.95 / 0.75 / 0.5). Matched entities surface their Q-ID + sameAs URL list (Wikipedia, official sites, social profiles).

What this gets you

• Is your client's site itself in Wikidata? This is a heavy AI Search ranking factor — LLMs (ChatGPT, Perplexity, Gemini) preferentially cite Wikidata-indexed entities
• Match rate of entities mentioned on your site → gives you a "knowledge graph fluency" signal
• sameAs coverage — how complete your schema's external-link declarations are
• Manual Q-ID override (Agency+) — bind a specific entity to a specific Q-ID if our auto-match misses it