Trustgent
About

Methodology

v1.0, published 2026-07-05. Every substantive change is logged in the quarterly notice-and-action report.

# Methodology

This page describes how Trustgent operates as a platform: how a buyer uses the directory to find an AI implementation partner, how a provider gets listed and moves through verification tiers L0 through L5, how ranking is computed under a fixed allowlist, what plan features providers can and cannot pay for, how editorial pages are produced and reviewed, and how disputes and corrections are handled. It is the procedure companion to the Trustgent Attestation Protocol and the verification mechanics.

How a buyer uses Trustgent

A buyer arrives at Trustgent with a concrete question: which AI implementation partner can deliver a specific capability, in a specific country or industry, at a verifiable level of demonstrated work. The platform is built to answer that question in three modes.

Mode 1: browse verified providers

The primary surface is the providers directory, filtered by capability (retrieval-augmented generation, agentic workflows, computer vision, forecasting, evaluation, MLOps, and so on), by country, and by industry served. Every listed provider carries a verification level between L0 and L5. Buyers can restrict results to L2 and above to see only cross-referenced providers, or to L4 and above to see only providers whose client work has been audited by Trustgent.

Each provider profile exposes the underlying record: which capabilities are attested, which client references corroborated which claims, which case studies are on file, and when the record was last verified. Nothing on a profile is editorialized in the provider's own voice. The profile shows the state of the record.

Mode 2: request verified matches (Phase 6, concierge)

For buyers who prefer a shortlist over a search, Trustgent will offer a concierge service that returns three verified matches within 24 hours based on a written brief. This mode is scoped for Phase 6 and is not yet live. When it launches, the matching function will use the same fixed ranking allowlist described below. It will not surface a paid tier over a higher-verified alternative. The matching output will include the reasoning: which verification records qualified each match, and which did not.

Mode 3: contact a provider directly

Every verified provider profile carries a direct contact route. Trustgent does not intermediate the commercial relationship, does not take a referral fee, and does not gate contact behind a paid subscription. The platform's role ends at the qualified handoff.

How a provider gets listed

Provider records move through six states. Each state has a defined evidence requirement and a typical processing time. The progression is one-way: a provider cannot skip a tier, and a provider that fails a higher-tier audit does not silently retain the lower tier's badge.

L0, claimed listing

L0 is a listing without verification. It is created either by the Trustgent discovery pipeline (which indexes public providers from company registries, industry associations, and public web sources) or by a self-serve claim from the provider. An L0 record shows the name, country, and public capability signals. It carries no badge and is not surfaced above L1 or higher records in any ranking. Typical time to reach L0 from discovery: immediate; from self-serve claim: same day.

L1, self-attested

L1 requires the provider to complete a structured profile: legal entity, registered address, capability declarations, industries served, team size, founding year, and named client references (with permission to contact). The provider signs a self-attestation that the information is accurate. L1 is a claim, not an independent verification. Typical time: one to three business days after claim.

L2, cross-referenced

L2 requires independent third-party corroboration of the L1 profile. The verification agent checks: registered legal entity matches a public company registry; named founders and technical leads have LinkedIn profiles consistent with the company record; capability declarations are consistent with the provider's public website, GitHub organization, published talks, and industry-body membership. Any discrepancy between the L1 claim and the third-party signal blocks the tier. Typical time: three to seven business days after L1.

L3, recorded proof

L3 requires client-authorized evidence of delivered work. A provider submits case study records: signed statement of work, redacted deliverables, client sign-off or written testimonial, and permission from the client to publish the case study on the provider's Trustgent record. The verification agent checks that the case study aligns with an L2 capability claim and that the client entity is verifiable. Typical time: two to four weeks per case study.

L4, audited

L4 requires a Trustgent-conducted structured interview with a named client reference, plus documented work outputs from that engagement. The interview follows a fixed protocol (see how we verify) and covers scope, delivered artifacts, outcomes against the original brief, and post-delivery support. The client reference is on the record by name. L4 is the highest tier available without ongoing monitoring. Typical time: four to eight weeks after L3.

L5, continuously monitored

L5 is L4 with a rolling audit cadence. The provider agrees to a freshness gate: at least one L4-grade audit refresh per year, plus verification-agent re-checks on public signals every quarter. If the freshness gate lapses, the record steps down to L4 until the next audit closes. L5 is a live state, not a permanent award.

How ranking works

Ranking is governed by a hardcoded allowlist. The allowlist is fixed in code as `RANK_ALLOWLIST` and contains exactly three fields:

1. `verificationLevel`, the dominant sort key. L5 above L4, L4 above L3, and so on. A higher-verified provider always ranks above a lower-verified provider in a comparable query. 2. `recordCount`, the tiebreaker within a tier. Within an L4 cohort, a provider with more L3 case studies ranks above a provider with fewer. Record count is a proxy for breadth of corroborated work, not for revenue or size. 3. `lastVerifiedAt`, the recency sort. Ties on level and record count resolve by most recent successful audit or re-verification.

Nothing else feeds ranking. Not plan tier, not brand assets, not ad spend, not referral bonus, not editorial preference, not the provider's geographic distance from the platform's operator, not the provider's language of primary operation. The ranking function is `ItemList`-shaped and every ItemList surfaced by the platform, including AI-answer-engine feeds, is passed through a filter (`aeoProviderItems()`) that strips any paid placement before the list is emitted. The firewall is enforced in code and covered by tests.

For a longer treatment of the ranking function, including the JSON output shape and the filter's behavior on edge cases, see how we rank.

What providers pay for

Trustgent monetizes plan features that do not touch ranking or verification. A provider on a paid plan gets:

  • Branding on the profile: custom logo, brand color, hero image, extended narrative section.
  • Additional seats: multiple team members can manage the provider record.
  • Analytics: profile views, capability-query impressions, referral-source breakdown.
  • Bulk invites: invite client references at scale for case study submissions.
  • Priority support: faster response on record updates and audit scheduling.

A provider cannot pay for:

  • A higher verification level. Verification is earned through the evidence chain above, not sold.
  • A better ranking position. Ranking is a function of the three allowlisted fields.
  • Suppression of a lower-verified competitor. There is no downranking-for-hire mechanism.
  • Editorial coverage in a capability explainer or buyer guide. Editorial pages are produced independently.
  • Removal of a public correction. Corrections are retained on the record per the corrections policy.

How editorial pages work

Trustgent publishes capability explainers, buyer guides, insights, and case studies under the Insights section. Each editorial page carries a named author, a named reviewing editor, a publication date, and a last-updated date. Where a page names specific providers, the provider is named on merit against the page's stated criteria, not against a paid arrangement. The editorial voice follows the editorial standards: neutral, precise, and without marketing superlatives.

Author and editor

Every editorial page names a human author and a human reviewing editor. Author bios list qualifications and disclosures relevant to the page's subject matter. The reviewing editor is accountable for factual accuracy and for adherence to editorial standards.

Sourcing

Editorial claims of fact are sourced. Where a claim references a verified provider record, the link points to the provider's Trustgent profile and, where applicable, to the underlying proof record. Where a claim references external material (a paper, a regulator's guidance, a public benchmark), the source is cited inline.

Corrections

Editorial pages are subject to the same corrections policy as verification records. A correction is dated, described, and retained on the page.

How disputes are handled

Two categories of dispute arise on Trustgent: a provider disputes its own verification record, and a third party disputes a claim on a provider's record or on an editorial page.

Both categories are handled through the process documented at corrections policy. In short: the disputing party submits the dispute through the platform's dispute channel, the verification agent (for record disputes) or the editorial team (for editorial disputes) opens a review, the disputed record's badge is suspended if the dispute is prima facie credible, the review runs against the evidence chain, and the outcome is either a confirmation of the original record, a correction on the record, or a downward tier adjustment.

Trustgent does not remove a well-founded correction because a provider objects to its visibility. The correction stays on the record. This is the trade-off that makes verification meaningful.

The verified badge is an editorial signal

The Trustgent verified badge is Trustgent's editorial signal that a provider's claims have passed the described evidence chain. It is not a regulatory certification and it is not a conformity assessment under the EU AI Act or any other statutory regime. Buyers who need statutory conformity for a specific regulated use case should consult a notified body or an accredited certifier for that regime. The badge is a procurement signal, not a legal instrument.

Related pages

Methodology v1.0, published 2026-07-05. Trustgent is operated by Rustenhoven Management B.V., KvK 92219500, Prinsengracht 234B, Amsterdam.