Independent opinion
Verdicts are SilentCritique editorial opinions based on the specific tests and evidence shown. They are not legal findings, regulatory determinations, or blanket claims about a company or person.
Editorial policy
The rules for public agent verdicts.
SilentCritique exists to help people decide which AI agents and machine outputs to trust. That requires direct public judgment, but it also requires restraint, evidence, reply rights, and a hard separation between money and outcomes.
Evidence-bound language
Published language should describe what was observed in the tested scenario. We avoid claims about unlawful, malicious, fraudulent, or bad-faith conduct unless independently reviewed and directly supported.
No dummy verdicts
The public third-party corpus must contain real tests only. Calibration examples are marked separately, kept out of search indexing, and never counted as Trust 100 outcomes.
Payment cannot buy outcomes
SilentCritique does not accept payment to remove criticism, change scores, suppress verdicts, delay publication, or award badges. If paid retesting is offered, payment covers evaluation labor only.
Disclose material relationships
A verdict should disclose whether the evaluation was unsolicited, requested, paid, sponsored, or affected by a material relationship with the subject.
Badge integrity
A badge must link to the canonical verdict page so users can inspect the date, methodology, evidence, limitations, status, and any published reply.
Pre-publication review
Before a third-party verdict is marked published, the page should pass a factual support check, a language risk check, a methodology check, and a right-of-reply readiness check.
Corrections stay free
Factual corrections, status updates, and right-of-reply requests do not require payment.
Read the reply process