Attested interpretive units

An attested interpretive unit is a selected canonical statement that Better Robots.txt publishes with enough metadata to be verified and scoped.

It is not the whole page. It is not a marketing slogan. It is a precise unit of meaning that should not be blurred during retrieval or summarization.

Required fields

Each unit declares:

• unitId;
• language;
• claim class;
• authority scope;
• canonical source;
• source path;
• selector;
• canonical text;
• allowed transformations;
• forbidden transformations;
• SHA-256 digest.

Why units are selective

The site does not hash every page.

It only attests units where integrity matters for interpretation, such as:

• product identity;
• policy-signal limits;
• /check audit boundaries;
• runtime evidence requirements;
• external criticism boundaries;
• hash limitations;
• llms.txt guidance boundaries;
• proof and case-study limits.

Current registry

The current registry is published at /attested-interpretive-units.json.

It is proposed, deterministic and audit-oriented. It is not a claim that third-party AI systems already use or respect these units.