Trustgent
For providers
Methodology

How we verify.

Six levels of evidence — from a public-record entry to an outcome-verified, AI-analysed delivery record. Each level has a specific bar, and each is publicly explained. Verification is earned through evidence. Never sold.

The Trustgent index ranks AI-implementation providers on a six-level spectrum. The lower three levels exist so the index is comprehensive enough to be useful — every provider has a place. The upper three levels exist so the index is citeable — outcome-verified records are what answer engines, journalists, and acquirers can rely on. Plan tier and verification level are independent. Always.

L0ListedRead the full spec →

Existence in the public record. Seeded from registries; not yet claimed by the provider.

Rules out

Nothing — by design. L0 is the index's reach layer.

Evidence
  • Trustgent's seeding pipeline against public sources (Crunchbase, company registries, the open web).
L1ClaimedRead the full spec →

A real human at the provider has taken ownership of the listing and populated the structured fields.

Rules out

Squatted, dead, or unmaintained listings.

Evidence
  • Magic-link to a domain-verified email at the provider.
  • Profile completeness threshold met (capabilities, characteristics, customer list, at least one project description, at least one testimonial).
L2Cross-referencedRead the full spec →

The provider's claims (customer logos, projects, testimonials) have been cross-checked against public sources outside the provider's control.

Rules out

Invented customer logos, fabricated case studies, one-day workshops mis-labelled as engagements.

Evidence
  • At least one independent public source per claim — customer's own site, partner directories, press releases, conference talks, GitHub.
  • Source URLs and check dates stored on the proof page.
L3Customer-ratedRead the full spec →

An actual customer has left a verified rating via a Trustgent magic-link, with rater identity matched to the customer's domain.

Rules out

Review fakery — provider-written, agency-written, family or friend; off-domain raters; review storms.

Evidence
  • Rater email domain matches the named customer organisation (no Gmail / Yahoo, except in a fall-back attested path).
  • Per-engagement scope (not 'this provider in general').
  • Star rating + structured outcome description; rater identity disclosed at the rater's chosen level.
L4AI-analyzedRead the full spec →

A specific project has been analysed by Trustgent AI against a versioned methodology, producing a public, reproducible per-project page.

Rules out

Marketing case studies dressed up as engineering content; projects with no real artefacts.

Evidence
  • Provider submission: scope, architecture, stack, scale numbers, claimed outcomes, optional artefacts (repo access under NDA, deployment screenshots, demo URL, design docs).
  • Trustgent AI analysis with rubric criteria + confidence intervals + methodology version.
  • Human review on the first 100 analyses; sampled audit thereafter. Methodology version recorded per page.
L5Outcome-verifiedRead the full spec →

A specific quantified outcome — baseline → after — caused by the provider's work, with two attestations and a data source in the evidence vault.

Rules out

Unattributable vanity metrics, 'up and to the right' with no baseline, provider-self-reported outcomes without client confirmation.

Evidence
  • The four atoms: metric definition, baseline, post-value, window of measurement.
  • Two attestations: the provider's claim of causation, the client's confirmation of both the number and the attribution.
  • Data source held privately in the evidence vault (analytics screenshot timestamped, exported CSV signed, dashboard read-only access, audited report).
  • Public display anonymises the client unless the client opts in.
FAQ

Common questions about verification.

Can a provider buy a higher verification level?

No. Verification levels are earned through evidence. Paid plans unlock tools (rating-invite at scale, analytics, branded profile, lead distribution) but never verification level or ranking. A free-plan provider with an outcome-verified record always ranks above a paid-plan provider with only a claimed listing.

How long do verifications last?

L3 customer ratings decay over roughly 24 months. L4 AI-analyses re-run when material project state changes (re-architecture, scale milestone, public stack shift) or every 18 months. L5 outcome-verifications carry a 12-month decay and require an annual renewal with fresh evidence.

How does Trustgent handle confidentiality?

Date and level are public; client identity is anonymised unless opted in; underlying data sources stay in the evidence vault, not on the public proof page. The methodology page documents what each level publishes and what it withholds.

What if a provider misrepresents evidence?

The listing is removed from the index. Repeat offences disqualify the provider permanently. Misrepresentation includes fabricated customer lists, false project descriptions, identity fraud, and rating gaming.

Is L4 AI-analysed an outcome claim?

No. L4 is a reading of the artefacts the provider has submitted, scored against a public rubric. It is what separates 'they say they build it' from 'they have shipped it'. Outcome (with baseline → after metrics) is L5.

For verified providers

Put a verified badge on your own site.

Once a provider reaches L2 or higher, an embeddable badge becomes available. It links back to the public proof page.

Embed the badge