← Back to home

How Trust Scores Work

Every score on AgentRank is built from verifiable signals — not self-reported data or paid rankings. Here's exactly what goes into the numbers.

Verified Developers

25%

Real companies with verified identities. No anonymous listings gaming the system.

Proof of Usage

25%

Reviews from users who actually used the agent. API logs, receipts, or verified connections.

Community Moderation

20%

AI-powered spam detection + community flagging. Fake reviews get caught.

Transparency Score

20%

Clear pricing, documented limitations, honest about what it can't do.

Response Rate

10%

Developers who engage with feedback rank higher. Abandoned agents rank lower.

Security & Compliance

10%

Third-party audits, SOC2, GDPR compliance. Verified, not self-reported.

Score → Grade

90–100A+
80–89A
70–79B
60–69C
50–59D
0–49F

Anti-Gaming Measures

  • Reviews require proof of usage — no anonymous one-click ratings
  • Velocity detection catches sudden review spikes
  • IP and account fingerprinting prevents duplicate reviews
  • All reviews are human-moderated before affecting scores
  • Paid plans don't affect scores — money buys visibility, not trust