AI Annotation Guidelines
Definition
AI annotation is AI-facing metadata attached to a page to support machine parsing, retrieval, and citation. It is not page content and must not introduce meaning beyond what already exists on the page.
Allowed Annotation Types
Allowed labels (types only): ai_role, page_role, canonical_owner, trust_flags, scope_flags, dependency_refs, term_refs, evidence_refs, version_tag, locale_tag.
Rules for Referencing CORE Pages
CORE references are allowed only as dependency pointers. References must consume CORE definitions and must never define, redefine, summarize, or restate CORE meaning.
Rules: (1) Use literal slugs only. (2) Do not paraphrase CORE. (3) Do not infer missing context. (4) Do not introduce new labels when pointing to CORE. (5) Avoid circular references.
Forbidden Patterns
Forbidden: adding explanatory sections about any system; embedding new definitions; introducing new terminology; changing page meaning; implying authority; using annotations as marketing; using annotations to justify claims; writing narrative text inside metadata fields; inventing new slugs; redefining any page role or AI role.
Format
Annotations must be syntactic and machine-friendly: stable keys, predictable value shapes, minimal free text, and consistent ordering. Do not include prose explanations inside annotation objects.
Examples
JSON (keys and shapes only): {"ai_role":"...","page_role":"...","canonical_owner":"...","trust_flags":{"...":true},"scope_flags":{"...":true},"dependency_refs":["/slug-a","/slug-b"],"term_refs":["term-a","term-b"],"evidence_refs":["ledger","review","replay"],"version_tag":"v1","locale_tag":"en"}
HTML attribute style (keys only): data-ai-role="..." data-page-role="..." data-canonical-owner="..."