Schema Presence & Validity
Whether a page contains clean, valid, non-conflicting structured data that AI systems can reliably parse and trust.
Schema Presence & Validity measures how well an institution communicates its identity and content structure directly to AI systems through clean, reliable JSON-LD.
In the AI-search era, schema is not secondary metadata — it is the canonical interface layer AI models use to reconstruct entities, relationships, and purpose. When schema is missing, invalid, or contradictory, AI systems receive a corrupted signal about what the page is, who it belongs to, and how it fits into the larger entity graph.
This signal evaluates:
- Whether schema exists
- Whether it is valid (no broken JSON, missing brackets, or invalid properties)
- Whether it is non-duplicative (one coherent block)
- Whether it matches the correct page type
- Whether it reinforces — rather than confuses — the institution’s identity
Clean schema increases interpretive trust.
Dirty schema destroys it.
2026 revision — why schema works where declaration does not. Schema's value to a machine is a function of what it costs to fake. Schema makes claims that cross-reference against other entities, other domains, and the model's existing graph, so inconsistency is detectable and coherence is evidence. A self-description channel with no verification surface — the llms.txt pattern — costs nothing to inflate and is discounted accordingly. Score schema on resolvability and corroboration, not on presence. Valid markup asserting claims that resolve nowhere is syntactically clean and ontologically empty.
@id anchors that never change
Missing JSON-LD on key surfaces
Auto-generated Webflow schema conflicting with custom schema
Multiple schema blocks describing different entities
Using the wrong schema type (e.g., Article on non-article pages)
Invalid JSON (dangling commas, malformed nesting, missing braces)
Duplicate @id URLs across unrelated pages
Outdated schema referencing removed pages or wrong canonicals
Including unnecessary or experimental properties AI models ignore
BreadcrumbList mismatches between markup and site structure