An independent registry · Provenance Standard v1.2Verify a mark · Report misuse · EN
ATC PROVENANCE STANDARD · VERSION 1.2 · EFFECTIVE 2026-04-01

The Standard

The published specification that defines what an AI Trust Certify mark attests, what evidence is required, and what the mark explicitly does not claim. Every certificate in the registry is issued against this document.

1 · Scope & purpose

The ATC Provenance Standard governs certification of a single piece of published content — a video sales letter, social ad, display creative, or landing page. It does not certify a brand, a company, a product, or a domain. A publisher may hold many certificates; each one covers exactly one asset and one set of disclosures made about that asset.

The Standard exists to make a small number of provenance facts about AI-generated and AI-assisted content independently checkable by the public. It is built on top of, not in competition with, the open C2PA Content Credentials standard for media provenance. C2PA establishes the chain of custody of a file; this Standard adds disclosure of AI involvement and verification of any featured licensed professional — the human-attestation layer that media-provenance alone does not cover.

2 · The three claims

A certificate may attest up to three independent claims. Each is recorded with its evidence source and a dated verification. A claim that cannot be supported is simply not asserted — the certificate shows only what was verified.

CLAIM 01 · DISCLOSURE

AI involvement

A factual disclosure of how AI was used in the asset, selected from a defined vocabulary: fully synthetic, AI avatar / presenter, AI voice / cloning, AI-assisted editing, AI-generated script, or none. This is self-declared by the publisher and recorded permanently. It is a label, not an endorsement, and requires no third-party judgment.

CLAIM 02 · PROVENANCE

Footage authenticity

Whether footage is genuine capture or synthesized. Where the publisher provides C2PA Content Credentials, the manifest is validated and the result recorded as Verified. Where authentic capture is claimed without a C2PA signature, it is recorded as Disclosed, unsigned — a weaker statement, shown honestly as such.

CLAIM 03 · CREDENTIAL

Licensed professional

If a licensed professional appears or is named, their license is confirmed against the issuing authority. For physicians this means the relevant state medical board (licensure) and ABMS (board certification), with the public record linked and a dated "verified as of." Licensure and board certification are distinct claims and are recorded separately.

3 · Evidence requirements

ClaimPrimary sourceRecorded as
AI involvementPublisher declarationDisclosed (self-declared)
Footage — signedC2PA manifest validationVerified
Footage — unsignedPublisher declarationDisclosed, unsigned
LicensureState licensing board public recordVerified, with link & date
Board certificationABMS / ABPS public lookupVerified, with link & date

Every verified claim links to the external source so a reader can repeat the check. The proof is never the issuer's word; it is the underlying public record.

4 · Validity & re-verification

Certificates are issued with an explicit issue date and expiry, and are re-verified on a 90-day cycle. Because a license can lapse or a disciplinary action can land, a static "verified" badge would become a false claim over time. If a re-check is not completed before expiry, the certificate moves to Expired automatically.

  • Valid — within its window and last re-verification passed.
  • Expired — re-verification window elapsed without renewal.
  • Revoked — withdrawn for cause (e.g. the mark was displayed outside its certified scope, or a credential failed re-check).

Expired and revoked certificates remain visible in the registry permanently. A public revocation log is itself a trust signal.

5 · What the mark is not

The mark certifies verifiable facts — AI disclosure, footage provenance, and licensure status. It is never a statement that the content is accurate, that the product works, that claims made in the content are true, or that AI Trust Certify endorses the publisher. Representing the mark as any of these voids the certification.

This boundary is deliberate and load-bearing. A narrow factual attestation that links to a primary source survives scrutiny even when the issuer has a relationship with the publisher, because the proof does not depend on the issuer's opinion. The moment a mark drifts toward "approved" or "trustworthy," it stops being defensible.

6 · Changelog

VersionDateChange
1.22026-04-01Added "Disclosed, unsigned" provenance state; separated licensure from board certification as distinct claims; formalized 90-day re-verification.
1.12026-02-15Adopted C2PA Content Credentials as the provenance backbone for Claim 02.
1.02026-01-10Initial standard: three claims, public registry, revocation log.