Suganthan: schema has three lives — index-time, training-time, query-time

Claim: Suganthan Mohanadasan argues schema operates in three layers:

1. Index-time — Google Knowledge Graph entity disambiguation
2. Training-time — canonical entity stores that feed next-generation LLMs (Wikidata, Schema.org corpus)
3. Query-time — real-time fetch by AI agents at the moment of a user query

Single-experiment studies like Ahrefs (April 2026): adding schema to 1,885 pages produced no AI-citation lift; AI Overviews showed -4.6% only measure layer 3. Layers 1 and 2 compound over years and account for most "schema is dead" hot-takes missing the point.

Source: Suganthan Mohanadasan, "The three lives of schema markup" — https://suganthan.com/blog/three-lives-of-schema-markup/

Confidence: Single-source synthesis (analytical framing, not empirical study).

Why it matters for Candid: It's the cleanest counter-frame to "Ahrefs proved schema is dead." Use when a client cites the Ahrefs result as a reason not to invest in schema. The honest position: Ahrefs measured one slice; the other slices accumulate value the experiment can't see.