Trustgent
About

Editorial standards

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

# Editorial standards

Trustgent's editorial contract is simple. Every page you read here is written to be checkable against a source, ranked plan-blind against a verified evidence chain, and free of promotional language a provider might have paid to place. These standards, versioned and dated, govern voice, sourcing, authorship, AI-assisted drafts, corrections, and conflicts. If a page ever falls short, the corrections policy tells you exactly how it gets fixed and on what timeline.

Voice: neutral, precise, no superlatives

Trustgent's editorial voice does not use "best," "top," "leading," "premier," "world-class," or any similar superlative to describe a listed provider. The reason is structural, not stylistic. Superlatives are the vocabulary of paid placement. When a directory ranks providers on merit but describes them in the language of an advertisement, the reader has no way to tell where verified fact ends and marketing begins. Trustgent removes the ambiguity by removing the vocabulary.

Neutrality is enforced in code, not policy. A build-time linter (`seo-lint`) scans every editorial file for a blocklist of promotional terms and fails the build if it finds one used to describe a provider. Pages that need to discuss the concept of "leading" or "best" in the abstract, for example when quoting a source, use a `neutral: true` flag in the page frontmatter that scopes the check appropriately. The default is strict.

The rank order shown to visitors is produced by `aeoProviderItems()`, a firewall function whose ranking allowlist is hardcoded to three fields: `verificationLevel`, `recordCount`, and `lastVerifiedAt`. No paid feature, whether branding, additional seats, analytics access, or bulk invites, can enter the sort. The provider is either verified or it is not. The record either exists or it does not. The last verification date is either fresh or it is stale. Everything else is out of scope for ranking, by design.

Providers are named on merit. Where a page lists provider names, the order is deterministic and reproducible from the public data. Where a page draws a comparison between providers, the comparison is grounded in fields you can inspect yourself on each provider's profile.

Authors and reviewers

Editorial pages are authored by the Trustgent Organization. Authorship is attributed at the organizational level rather than to individual bylines, and the schema.org author entity on every article is the Trustgent Organization. This matches how the underlying methodology works: the ranking is plan-blind and reproducible from public inputs, and the editorial pipeline that surfaces it is the organization, not any one person.

Reviewers are a separate role. Where a page describes a technical process, a verification tier, an audit methodology, a scoring formula, the reviewer is a domain expert in the relevant field: procurement, information security, machine-learning operations, or corporate law. Reviewer identity is disclosed at the foot of the page along with the review date. Reviewer names are omitted from JSON-LD to keep authorship attribution consistent at the organizational level; the on-page footer is the record of who reviewed what and when.

Where a page has no separate reviewer, it was reviewed against these standards by the Trustgent editorial process and the reviewer field will read "self-reviewed." This is disclosed rather than hidden.

Sourcing standards

Direct sources outrank secondary reporting. If a claim can be traced to a primary document, for example a company registration, a research paper, a court filing, or a provider's own case-study documentation, that primary document is linked. Secondary sources (trade press, blog write-ups, industry-analyst notes) are cited only when the primary source is not public, and only alongside a note explaining why.

Every non-trivial factual claim links to its source at the point the claim is made, not in a bibliography at the bottom. If a page states that a provider has completed a given number of engagements, that number is either grounded against a specific verified database row (with the row's public identifier visible) or it is not stated at all. Trustgent would rather leave a claim off the page than support it with a source the reader cannot inspect.

Opinion is labeled as opinion. Where the editorial voice makes an inference, for example "this is why the industry is moving in a given direction," the paragraph is prefixed with a note that this is analysis rather than a sourced fact. Analysis paragraphs are held to a lower evidence bar than sourced claims but to a higher bar of internal consistency. A claim of analysis must follow from something already on the page.

Provider data claims are always grounded against the verified database, not against provider marketing. If the provider's website says one thing and the verified record says another, the verified record wins in editorial. If the divergence is material, the provider is invited to submit evidence through the corrections process. See the Trustgent Attestation Protocol for the evidence chain that governs what counts as a verified fact about a provider.

AI-assisted content policy

Some Trustgent content is drafted with the assistance of large language models. This is disclosed here in full and, where material to a specific page, at that page's footer.

Where AI assistance is permitted:

1. Structured content generated from verified database rows: provider summaries, cluster hub blurbs, capability tables, country facets. The language model is acting as a formatter for facts that already exist in the verified database. 2. Initial drafts of buyer guides where the structure is templated and every factual claim resolves to a linked source or a verified row. 3. Copy-editing passes on human-written prose, limited to grammar, structure, and readability. The model is not permitted to introduce facts, examples, or claims during a copy-edit.

Where human authorship is required:

1. Editorial essays and analysis pieces where the argument is the value. 2. Methodology posts, including this one, that describe how Trustgent's own systems work. 3. Any content that attributes a position, motive, or intent to a named person or organization. 4. Corrections notices and any post-publication revision to the substance of a page.

AI-assisted content is labeled in compliance with Article 50 of the EU AI Act. Where a page's substantive prose was drafted by a language model and reviewed by a human editor, the page footer carries a machine-readable label indicating the AI-assisted status. Where a page is fully human-drafted, no label is emitted. The absence of the label is meaningful.

The distinction Trustgent draws is between generation and inference. A language model turning a verified provider record into readable prose is generating from a grounded source. A language model inferring a fact that is not in the source is not permitted in Trustgent editorial. That class of failure is treated by the corrections process as a factual error, not as a stylistic issue.

Corrections and updates

The corrections policy is published in full at /about/corrections-policy. In summary: any reader can submit a correction request, corrections are triaged within two business days, factual corrections are executed as soon as the correction is verified, and corrected pages carry a visible correction notice for thirty days after the change lands.

Update dates are substantive-change based, never touch-based. A page's "last updated" field is only advanced when the page's substantive content changes: a fact revised, a section added, an argument tightened, a claim retracted. Reformatting, link maintenance, and typo fixes do not advance the date. This means the update field is meaningful. Readers can rely on it to know whether the page has been meaningfully re-examined since they last read it.

Versioned documents (this page, the methodology, the how-we-verify page, and the Trustgent Attestation Protocol) carry an explicit version number in the footer. When a versioned document changes materially, the version increments and the change is logged in the document's history.

Independence and conflicts

Trustgent is backed by Rustenhoven Management B.V., a Dutch entity holding equity in the platform (KvK 92219500, registered at Prinsengracht 234B, Amsterdam). The platform is not owned by, invested in, or paid by any provider listed on Trustgent.

Providers pay Trustgent for plan features on their own listings: custom branding, additional user seats, bulk invitation flows, analytics dashboards. Providers do not pay for ranking, do not pay for inclusion, and do not pay to influence editorial coverage. This is the plan-blind moat, and it is enforced in code. See the how-we-verify page and the Trustgent Attestation Protocol for the mechanism that keeps paid features out of the rank function.

If the operator or a named Trustgent author holds equity, a board seat, or an advisory role at a listed provider, that relationship is disclosed at the top of any page that mentions the provider, and the affected provider is excluded from any editorial recommendation list on the page. As of the version date on this document, no such relationships exist.

Trustgent's verified badge is an editorial signal grounded in Trustgent's own methodology. It is not a regulatory certification. It is not a conformity assessment under the EU AI Act or any comparable statute. Providers displaying a Trustgent verified badge are asserting the outcome of Trustgent's audit process, not compliance with any specific regulation. Readers who need regulatory conformity should look to notified bodies and accredited certification schemes. Trustgent's role is orthogonal: to make procurement of AI implementation work more accountable to evidence than to marketing.

Related pages

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