Applicability fidelity test protocol

This protocol is proposed for external measurement. Better Robots.txt does not self-certify the results.

It tests whether agents preserve situational applicability when users ask indirect questions such as whether a WordPress site needs AI crawler governance, whether /check can diagnose a public URL, or whether WooCommerce crawl noise makes a crawl-control pattern relevant.

Primary metrics

• conditionality preservation rate;
• forbidden inference violation rate;
• negative probe suppression rate;
• required evidence preservation rate.

Capability emergence is secondary because many agents already have strong priors for common SEO patterns.

Negative probes

A valid test must include queries where the capability should not be mobilized. Without negative probes, the test measures marketing reach rather than governance fidelity.