This lesson runs a whole site’s entity list against Google’s Knowledge Graph and turns the result into the Connection score. It is as much about reading the output honestly as running it: most of the value is in knowing which verdicts to trust, which to reject and which to leave alone.
What this lesson covers
- The Reconciler’s six steps, from upload to export
- The match-score floor and the lookup cap
- Matched, ambiguous and unmatched, and what each means
- Type conflicts hidden inside the matched count
- Near-name collisions, type mismatches and genuine ambiguity
- Scoring Connection, with the brand scored first
The Knowledge Graph Reconciler takes the entity export from Module 2, de-duplicates it to distinct entities, looks each one up against Google’s graph and keeps the best candidate. The lesson explains why the export matters more than a typed list: the type column unlocks type-conflict detection and the MID column unlocks over-linking detection in Lesson 3.4. It sets the match-score floor from the demo corpus — twenty, the lowest score among genuine matches — and shows how to derive your own rather than inherit it. Lookups are free up to 100,000 a day, so the whole lesson costs nothing to run.
On the demo site, 5,000 entities split into 2,118 matched, 776 ambiguous and 2,106 unmatched, with 149 type conflicts hiding inside the matched figure — confident matches to the wrong kind of thing. The lesson argues that those conflicts, not the unmatched share, are the number to take to a client, because most unmatched entities are ordinary nouns that were never going to be graph entities. It also asks you to resist the temptation to promote ambiguous entities to matches: an invented identifier propagates through every later step. Connection is then scored with one rule that stops it being gamed — if the brand does not resolve, the score cannot rise above one, however many generic entities match.
This lesson continues Module 3, Knowledge Graph Auditing & Entity Mapping, in The AI-powered Entity Audit, and is followed by a step-by-step practical on the Reconciler. It suits SEOs who need to report how well a site’s entities connect to Google’s graph without overstating either the gaps or the matches.