Does the engine know which entity this is?
Confirm the entity resolves to one unambiguous node — a common name with no disambiguating markup gets conflated with anyone else who shares it.
GEO pathway · foundation 01
Ranking is not the same problem anymore.
A generative engine doesn't return ten links for a human to sort through. It resolves an entity, retrieves what it can find, and writes a paragraph as if it were fact. Whether that paragraph is correct depends on conditions that have nothing to do with keyword position.
Learning boundary. This lesson teaches an operating model. It is not a guarantee of ranking, citation, or inclusion in any specific engine's output.
Learning objective
Separate five legibility conditions.
By the end of this lesson, you should be able to inspect a given entity's web presence, assign evidence-based readiness scores across five conditions, identify which gaps cause hallucination versus which cause invisibility, and explain why dense content is not the same as machine-legible content.
Is the ground truth machine-readable?
Publish facts as parseable schema, not just prose — a fact stated only in a paragraph requires the model to infer structure that explicit markup would have handed it directly.
Can a crawler or RAG pipeline actually reach it?
Verify the content is indexable, uncrawled-blocked, and present in the sources retrieval engines actually query — authoritative content behind a wall an engine can't cross doesn't exist to it.
Is the source traceable and authoritative?
Anchor claims to a canonical, citable source — an engine synthesizing from unattributed or low-authority mirrors has no way to weight your version as ground truth over a stale copy.
Is this still true, and does the engine know that?
Keep dated facts current and re-published — a five-year-old snapshot with no update signal gets presented as the present state, because nothing told the model otherwise.
Practice lab · evidence before confidence
Score an entity's AI legibility.
A founder's official bio page states their current role, company, and background in well-written prose. There is no Person schema, no sameAs linking to their other verified profiles, and the page has not been updated since an earlier company they founded was still their primary venture. A Google Knowledge Panel does not exist for this person.
- Score each dimension from 0: absent, 1: documented, or 2: tested.
- Name every claim that requires evidence before awarding a point.
- Identify the critical gaps that prevent reliable AI-generated description.
- Write the smallest safe next experiment.
Answer rationale
Structured data is 0 — well-written prose is not machine-readable structure, regardless of how clearly a human would understand it. Resolution is at most 1: without schema or a knowledge panel, nothing forces disambiguation if another person shares the name, so the risk is documented by its absence, not tested. Attribution is 1 at best — the bio is presumably authoritative but isn't reinforced by sameAs links an engine could use to confirm it against other sources. Currency is 0: the page actively presents outdated information as current, which is worse than silence. The appropriate next step is adding Person schema with sameAs links and republishing the bio with current facts — not writing additional prose.
STOP → STRUCTURE