Skip to content

AI-Retrievable Standards

Source version 1.2.0. Draft for discussion and validation.

12. AI-Retrievable Standards Overlay

12.1 Purpose

This overlay makes standards content reliably retrievable, interpretable and citable by AI-assisted search and retrieval-augmented generation systems without prescribing a particular model, vector database or vendor.

12.2 Requirements

  • stable identifiers at document, clause, table, rule and concept level;
  • explicit document hierarchy;
  • normative/informative classification;
  • language and jurisdiction metadata;
  • effective version and applicability;
  • machine-readable definitions and references;
  • retrieval units aligned to semantic boundaries rather than arbitrary token windows;
  • citations back to authoritative identifiers;
  • supersession and deprecation relationships;
  • integrity hash and publication timestamp;
  • access and rights metadata;
  • examples and counterexamples separated from normative requirements;
  • links to schemas and SHACL shapes.

12.3 Retrieval result contract

A retrieval response should be able to return:

  • the authoritative text or structured assertion;
  • component identifier;
  • parent standard and version;
  • normative status;
  • effective date;
  • definitions required for interpretation;
  • referenced components;
  • provenance and validation status;
  • canonical human-readable link.