Attested integrity layer
The attested integrity layer is a proposed governance layer for Better Robots.txt.
It answers a narrow question: did a machine reader use the same canonical unit that Better Robots.txt published?
It does not prove that the unit is true. It does not prove crawler obedience. It does not prove that a generated summary is faithful.
What the layer attests
The layer publishes selected canonical units with:
- a stable unit identifier;
- a claim class;
- an authority scope;
- a canonical source;
- canonical text;
- allowed transformations;
- forbidden transformations;
- a SHA-256 digest.
The digest is computed over the canonical unit envelope, not over a weak scalar such as true.
Why this matters
Better Robots.txt repeatedly needs to preserve distinctions such as:
- policy signal versus enforcement;
/checkscore versus crawler compliance;- llms.txt guidance versus ranking guarantee;
- documented capability versus active user-site state;
- official position versus independent evaluation.
A digest helps an auditor verify that the exact canonical unit was available and intact.
What the layer cannot prove
A hash cannot prove:
- truth;
- legal force;
- reputation;
- model compliance;
- external acceptance;
- crawler obedience;
- summary fidelity.
Those require evidence, source qualification or behavioral evaluation.
Machine-readable artifacts
Primary artifacts:
Relationship to interpretive weighting
Attested integrity says whether a canonical unit is intact.
Interpretive weighting says how much that unit should matter for a specific class of claim.
Both layers are needed. Neither one replaces direct evidence for runtime crawler behavior.