Cross-Page Entity Resolution in the Perception Graph
The same entity can appear on dozens of pages across a domain. Cross-page entity resolution determines whether the AI system treats these as the same entity or fragments them into multiple conflicting representations.
Entity resolution is the process of determining whether two entity mentions — on different pages, in different contexts, with possibly different phrasings — refer to the same real-world entity. When an AI system crawls a domain, it encounters the primary business entity on the homepage, the about page, the blog, product pages, case studies, and the schema markup on every page. Cross-page entity resolution is what turns these many mentions into one coherent entity model.
Resolution Signals
Cross-page entity resolution uses four types of signals. Name matching: variations of the entity name (full name, abbreviation, shortened form) are normalized and matched. Attribute consistency: when two entity mentions share the same founding date, location, and category, they are likely the same entity. sameAs linking: when two pages both carry the same sameAs URL pointing to an external identity source, they are definitively describing the same entity. Schema @id matching: when multiple schema blocks carry the same @id value, they are explicitly declared as the same entity.
Resolution Failures
Cross-page resolution fails when: the entity name is used inconsistently across pages (full name on one page, abbreviation on another, different capitalization on a third — all without a normalizing schema declaration), the entity attributes conflict across pages (different founding dates in different pages), or no sameAs or @id provides a stable external anchor. When resolution fails, the AI system may create multiple separate entity records for the same real-world entity, fragmenting its understanding of the domain.
▲The most common resolution failure pattern: a brand name appears in full on the homepage and about page, but only as a shortened form or abbreviation in blog posts and product pages. Without schema to declare equivalence, some AI systems treat these as different entities. The fix: add a sameAs or @id to the schema on all pages, and use consistent naming in body text.
Resolution in SiteNexis Entity Analysis
The Entity Intelligence Engine performs cross-page resolution as part of entity normalization. All entity mentions across the crawled pages are normalized by name, type, and attribute overlap. Conflicts are flagged as EntityIssues with descriptions of the specific attribute that conflicts and which pages carry each version of the attribute. The resolution output feeds directly into the Perception Graph construction; only resolved (normalized, deduplicated) entities become perception_nodes.
Perception vs Fact Layer — Part 10 of 10