Methods and boundaries
What does Hide’s evidence actually prove?

Ask about a method, source or limitation. Hide distinguishes an observation from a verified outcome and an unknown from a pass.

General information from an AI assistant, not a real check. Private by design, never sold or used for training.

Clear benchmark

Clear benchmark results

This public benchmark reruns the labeled edge cases used to test Clear's verdict logic. It provides reproducible evidence about known cases and does not provide independent certification or universal accuracy.

Current test result

Method
clear-public-v5
Labeled cases
105
Exact matches
105 of 105
False positives in this set
0
False negatives in this set
0

Download the machine-readable result

Checks included

The set includes common provider records, malformed and contradictory policies, null MX handling, network-address safety, domain normalization, MTA-STS policy coverage and DNSSEC-backed SMTP DANE parameters.

Labeled cases by check kind. n = 105.
Check kindCases
mx6
spf16
dmarc14
mta-sts7
mta-sts-policy9
tls-rpt7
dane-tlsa7
caa8
address22
domain9

Checks not yet measured

Production retry rate is not published as a number because a stable real-world sample and minimum observation window have not yet been reached. Real-world false-positive and retry rates require permission-safe reviewed samples, stable traffic and a declared observation window. Hide will publish those separately instead of presenting synthetic cases as customer outcomes.

Result interpretation

A match means the current method produced the labeled outcome for that case. It does not prove that every DNS provider, outage mode or live mail-server configuration is represented. Method changes are versioned, and the benchmark must move with them.

Review the method changelog