This lesson completes the module by asking whether the rest of the web — and the models trained on it — confirm that your brand exists and is what you say it is. It produces the fifth and final score, and a backlog of contradictions to fix.
What this lesson covers
- What independent corroboration is and why your own site is weak evidence
- The checklist: Wikidata, Wikipedia, consistent naming and profiles
- A cold-knowledge test on a model, run so it measures something
- The difference between parametric knowledge and retrieval
- Scoring Corroboration
- The corroboration backlog and how to re-test
A site asserting facts about itself is the weakest evidence there is, because every site does it. Entity resolution weighs independent, structured, repeated confirmation: a Wikidata item, a Wikipedia article where notability allows, consistent naming and descriptions across third-party profiles, and citations that use the canonical form. The lesson sets out a checklist in that order and explains why Wikidata comes first — it is structured, has no notability bar comparable to Wikipedia’s, and is the cheapest structured corroboration available to a small brand.
The second half is the test everyone wants to run and most run wrong: asking a model about your brand. With web search genuinely off and explicit permission to say it does not know, a model asked about the demo brand and its founder recognises neither — an honest failure, and a measurable one. Yet AI search with retrieval surfaces the brand 17 times. The lesson separates resolution, memorisation and retrieval so each can be stated precisely, scores Corroboration at one out of three on the demo site, and turns every contradiction between the site, the graph and the model into a backlog row.
The lesson also explains why Wikipedia should not be treated as a growth tactic: its notability bar is the reason it carries weight, and an article about a brand that does not meet it is likely to be deleted in a public discussion. For most small brands, a well-maintained Wikidata item and consistent third-party profiles are the realistic route.
This lesson completes the core teaching of Module 3, Knowledge Graph Auditing & Entity Mapping, in The AI-powered Entity Audit, and is followed by a practical on running the corroboration checks. It suits SEOs and brand teams who are asked what AI “knows” about them and need an answer they can defend.