The answer looked convincing
We were researching an SEO professional named Scott Allen. Search surfaced another SEO professional with the same name, relevant keywords and a stronger indexed footprint. The biography was coherent. It was also the wrong human.
That is the dangerous version of an AI-search error: not gibberish, but a fluent assembly of individually plausible facts attached to the wrong entity. If published, it would have reinforced the collision across another domain and made the next system’s job harder.
Entity work starts with evidence
The correction was not “add more keywords.” It was to establish authoritative anchors: the intended LinkedIn profile, the intended personal domain, first-party career text and client evidence supplied by the subject. Then we excluded every claim whose source path led back to the namesake.
Schema.org defines sameAs as a URL that unambiguously indicates an item’s identity. Google says Organization structured data can help disambiguate an organization, but markup is an aid to understanding, not permission to state uncertain relationships.
Build an identity evidence table
For every claim, store the entity, claim, source URL, source type, date observed, confidence and publication permission. Separate first-party evidence from third-party corroboration. Names, employers, client relationships, dates and quantified results deserve independent rows.
Use stable identifiers wherever possible: canonical profile URLs, domains, organization pages and consistent author bylines. Avoid copying a full graph onto every property. Each site should publish only the relationships relevant to its purpose, using stable @id values and verified sameAs references.
Structured data cannot rescue contradictory visible content. Nor does JSON-LD “trigger rapid indexing.” Google notes that even technically compliant pages are not guaranteed to be crawled, indexed or served in its explanation of how Google Search works.
The lesson for AI visibility
Entity recognition is partly an information-quality problem. Clear first-party biographies, consistent names, factual About pages, corroborating professional profiles and appropriately scoped schema reduce ambiguity. Distinctive proof—projects, specialties, geography and outcomes—helps separate people who share a name.
The same discipline applies to local presence. Do not create a Google Business Profile or borrow an address simply to manufacture proximity. An online-only business is not made locally eligible by hiding a residential address. Entity work that begins with a fiction creates suspension risk and contaminates the very trust signals it was meant to improve.
Optimization comes after identification. Before asking about AI visibility, ask whether the system knows which brand—and whether every fact in the answer belongs to the same real-world entity.