This lesson works through the entities the Knowledge Graph does not recognise and decides what, if anything, to do about each. It then turns those decisions into the artefact that keeps a site consistent — a canonical naming table with an owner — and produces the Consistency score.
What this lesson covers
- Triaging unmatched entities into three piles
- Telling a Knowledge Graph problem from a corpus problem
- Choosing between building recognition, renaming and dropping
- Finding naming drift with the Entity Consistency Checker
- The canonical naming table writers will actually use
- Scoring Consistency
Every unmatched entity is one of three things: genuinely novel, such as a brand or a named method; badly named, real but too generic for any resolver to pin down; or extraction noise that slipped through the filters. Almost everything lands in the third pile, and the lesson’s most useful diagnostic is that a large noise pile means going back to the Module 2 filters rather than starting an entity programme. Of what remains, building recognition is expensive and should apply to the brand and very little else, while renaming is underrated: on the demo site, the bare “MLforSEO” is typed as a person 102 times in 148, while “MLforSEO Academy” is an organisation in all 92 of its mentions.
The Entity Consistency Checker then measures whether one name is used everywhere. On the demo site it groups more than two hundred entities named in more than one way, 76 of them high severity because Google’s identifier shows they are the same thing. The lesson walks through real clusters to show that not every variant is an error — “Google” and “Google Cloud” share an identifier but are different things — and ends with a mapping table of entity, canonical name, accepted variants, MID and owner, placed where writers already work. The demo site scores one out of three, with 28 surface forms of the brand and three different types.
This lesson continues Module 3, Knowledge Graph Auditing & Entity Mapping, in The AI-powered Entity Audit, and is followed by a practical on the Consistency Checker and the naming table. It suits SEOs and content leads who need one agreed name for each important entity and a way to keep it.