Trustgent
methodology

What we count as "delivered", six verification levels.

Listed. Claimed. Cross-referenced. Customer-rated. AI-analyzed. Outcome-verified. Six levels of evidence, each with a different bar.

Most directories rank providers on how visible they are: ad spend, review volume, the size of a sales team. Trustgent ranks on evidence of delivery. To do that fairly we had to define, precisely, what counts as proof, and how much weight each kind of proof carries. That is the six-level verification spectrum.

The six levels

The spectrum runs from a provider simply existing in the index to a third party confirming that a build actually worked.

  • L0, Listed. The provider is in the index, assembled from public information. No claim has been made by the provider themselves. This is a starting point, not an endorsement.
  • L1, Claimed. Someone at the company has verified they control the listing (via a domain-matched email). The profile is now maintained by the provider, but nothing about their work has been checked.
  • L2, Cross-referenced. Specific claims on the profile have been checked against an independent public source the provider does not control, a case study on a client's site, a conference talk, a regulatory filing.
  • L3, Customer-rated. A real customer of a real engagement has rated the work. Ratings come through an invite tied to an actual project, not an open form anyone can fill in.
  • L4, AI-analyzed. A specific project has been analysed against a published methodology: what was built, how it was evaluated, whether it reached production. This is analysis of the build, not a claim about its business outcome.
  • L5, Outcome-verified. The strongest bar. Both the provider and the client have attested, independently, that a defined engagement closed and delivered. This is the only level that asserts an outcome.

Why the levels are ordered this way

Each step up the spectrum removes a way the previous level could be gamed. A claim (L1) can be self-serving, so L2 checks it against a source the provider doesn't own. A cross-reference still reflects the provider's framing, so L3 brings in the customer. A single rating can be cherry-picked, so L4 examines the engineering of an actual build. And analysis can still flatter a demo that never shipped, so L5 requires both sides to attest that the work closed.

What the levels are not

The spectrum is earned, never sold. A provider cannot pay to move up it, and paid plans do not change a level by a single notch. Ranking reads the level and the records behind it, it never reads who paid for the channel that lets a provider receive buyer briefs. That separation is enforced in code and tested on every change.

A level is also not a verdict on a company as a whole. It describes the evidence attached to specific records, a cross-reference here, a rated engagement there. A firm doing excellent work that hasn't yet gone through verification sits low on the spectrum not because the work is weak, but because the proof hasn't been gathered.

That is the trade we ask buyers to make with us: trust the evidence, not the volume of the marketing. The levels are how we make the evidence legible.

Newsletter

Stay ahead of the AI services market.

One email a month: what's actually being delivered, verified outcomes, rate benchmarks, AI-analysed builds, category shifts. No vendor PR.

By subscribing you agree to our privacy notice. Unsubscribe in one click at any time.

In this cluster

AI verification & trust