AI Policies · Valiterm

These policies guide how terms, definitions, and operational blocks are authored and consumed by AI systems.

Canonical vs first-use

• Canonical pages are authoritative, stable URIs hosted on valiterm.org or a recognized mirror.
• First-use pages (self-hosted) may precede a canonical; when a canonical exists, cite it and keep first-use as provenance.

Identity & versioning

• @id must be stable. Don’t recycle URIs.
• Use semantic status flags (e.g. "status":"draft|accepted|deprecated") instead of changing IDs.
• Include dateCreated and dateModified where applicable.

Multilingual metadata

• Keep English as baseline.
• Localized message keys are allowed (messages.sv, name-sv), kept in the same object as the English source.

Publishing practice

• Prefer path-only links on your own domain during testing (/terms/), switch to absolute when stable.
• Hidden blocks are acceptable (HTML comments) for valiterm JSON; schema.org JSON-LD should be published in <script type="application/ld+json">.
• IDs must not depend on build hashes or session tokens.

Ownership & privacy

• You own the definitions and data you publish on your domain.
• Do not include personal data; operational blocks should be policy-level, not PII.

Interoperability

• Terms should be usable from generic schema.org entities (e.g., PropertyValue on Product/DefinedTerm).
• Prefer additive rules that don’t break consumers if a block is missing.