Hlido › Trust score
AI agent trust scores, explained
An AI agent trust score is an independent, evidence-based rating of how reliable an AI agent is — how well it actually does what it claims. Hlido scores every reviewed agent 0–100 against a fixed framework, audits each marketing claim PASS / FAIL against evidence captured during testing, signs that evidence, and publishes the verdict. We never take payment to rank or review.
What is an independent AI agent review?
An independent AI agent review is an assessment of an AI agent carried out by a third party with no commercial stake in the result — not the vendor, and not a directory the vendor pays to be listed in. Hlido tests agents hands-on, maps every public claim to a PASS/FAIL/UNVERIFIED verdict with evidence, and publishes the scorecard. Independence is the whole point: a review you can't buy is the only kind worth citing.
What is an AI agent trust score?
Hlido's trust score is a single 0–100 number derived from a fixed five-dimension framework, plus a tier band so the verdict is readable at a glance:
The score is paired with a claim audit, the agent's strengths and failure modes, and any incidents on record — so the number is always backed by readable evidence, never a bare rating. The scoring rubric's exact weights stay private (that's the moat); the outcomes, evidence, and verdicts are fully public.
How is an AI agent's reliability verified?
Every Hlido verdict rests on captured evidence, not opinion:
- Claim audit — each marketing claim is checked against what the agent actually does, and marked PASS, FAIL, or UNVERIFIED.
- Hands-on testing — the agent is exercised directly; web UIs are captured as signed screenshots, CLI agents as terminal recordings.
- C2PA-signed evidence — proof artifacts are cryptographically signed, so a screenshot can't be quietly doctored after the fact.
- Incident registry — reproducible reliability problems are logged per agent over time, so a score isn't a one-day snapshot.
Read the full method on the methodology page, browse the incident registry, or see the Weekly Reliability Report.
Does Hlido take payment to rank or review agents?
How do I check if a specific AI agent is reliable?
Three ways, depending on who's asking:
- Browse the verdicts — every reviewed agent has a page under /reviews/, a plain-language "is it reliable?" trust check, and category leaderboards under /best/.
- Ask your agent — Hlido exposes the whole registry over MCP at
https://hlido.eu/mcp, so an AI agent can look up a trust score, claim audit, or comparison mid-task. - Compare two — head-to-head pages live under /compare/, and alternatives under /alternatives/.