entity-registry

Run an entity audit that checks six signal categories: structured data, knowledge bases, NAP+E consistency, first-party content, third-party corroboration, and AI recognition. Each observation is recorded with source, date, and evidence type, then summarized into a prioritized action plan.

Create a Wikidata entry with referenced statements, add Organization schema with sameAs links on your site, and build authoritative third-party mentions. Knowledge Panels typically follow once the Knowledge Graph has consistent, verifiable entity signals.

The entity registry owns machine-facing identity: canonical type, aliases, QIDs, sameAs links, and disambiguation evidence. The narrative registry owns human-facing canon such as positioning, messaging, voice, and approved descriptions.

No merge happens on similarity alone. Similar names, logos, domains, or descriptions are insufficient; a verified cross-link or explicit user confirmation is required before reconciling duplicate entity IDs.

Unknown means a tool or engine could not be observed, while Partial means evidence exists but is incomplete. Conflating them would misrepresent audit coverage, so unavailable tools are always recorded as Unknown, never as Fail.

Person records require an applicable lawful basis before persistence, use pseudonymous aggregate IDs, and exclude raw email, phone, and address data. A prior erasure or tombstone blocks recreation until the user explicitly authorizes a new record.