Do AI systems fetch, parse, and return to this property? Measured from observable request behavior and structural completeness, not from citation claims.
Every signal in this dimension is deliberately conservative. Fetch behavior and endpoint coverage are observable in server logs. Citation, answer inclusion, and training use are not observable from here, so they are not measured, no matter how much more impressive they would sound.
The underlying position: structure makes a property legible, and legibility is a precondition for being used rather than a cause of it. Anything claiming otherwise is selling something.
The proportion of classified crawler requests coming from AI vendors rather than traditional search crawlers.
Measured. Vendor identification relies on user agent claims supported by published address ranges, and both carry uncertainty.
Whether machines actually fetch the property's machine-readable endpoints, or only its pages.
Measured. Notably, this is one of the few checks that can prove structural work reached anything at all.
The share of pages emitting valid, type-appropriate structured data with real cross-references.
Measured directly and repeatable at any time.
Whether the machine-readable endpoints currently return valid, current, non-contradictory data.
Measured. Checked at build and worth re-checking after any content push.