From Link Graph to Fact Graph: How Internal Links Become Machine Facts
The internal link graph is a human-centric concept: pages connecting to pages. The Fact Graph is a machine-centric concept: typed, verifiable relationships between entities. Here is how SiteNexis translates one into the other.
The internal link graph is a page-level structure: node for each page, edge for each link between pages. The Fact Graph is an entity-level structure: node for each entity, edge for each verifiable relationship between entities. Both are derived from the same source — the crawled DOM — but they model different things. The link graph models the site's information architecture. The Fact Graph models the site's knowledge architecture.
The Translation Process
Translating the link graph into fact graph entries requires the entity data from each page. When a page links to another page with anchor text "our machine trust methodology," the translation produces: a fact_edge from the entity of the source page to the entity of the target page (or to the page itself as a node if no entity is clearly identified), with relationship type inferred from the anchor text and surrounding context, and an evidence_url pointing to the source page.
What the Fact Graph Adds That the Link Graph Cannot
The link graph shows that page A links to page B. The Fact Graph shows that Organisation X offers Service Y (when page A is the organisation's about page and page B is a service page, and the anchor text explicitly names the service). The semantic richness of the Fact Graph relationship — its type, its evidence, its entity grounding — is what makes it useful for AI trust analysis. The link graph is a structural map. The Fact Graph is a semantic knowledge graph.
●In the SiteNexis dual-write architecture, fact_edges are written simultaneously from two sources: the link graph analysis (which contributes all body links as structural edges) and the entity intelligence analysis (which enriches edges with entity grounding, relationship type, and confidence). The resulting fact_edge is richer than either source alone.
Using Both Together for Diagnosis
The most powerful diagnostic use of both graphs together: identify pages with high internal PageRank (link graph signal) and low entity confidence (Fact Graph signal). These are hub pages that the link architecture treats as authorities, but that have weak entity grounding — the AI system will discover them through link traversal but will not build strong entity knowledge from them. The fix is always content and schema, not more links.