Google Imagen check: reading the credential, not the pixels
Imagen output travels through the same provenance machinery as the rest of Google's generative stack, so the useful evidence is a C2PA manifest rather than anything visible in the picture. AICheck365 parses that manifest in your browser and reports the signature state, the trust state and the declared source type separately — because a credential that exists is a different finding from one that verifies.
Short answer: check digitalSourceType, then check the certificate
Look for a manifest declaring digitalSourceType = trainedAlgorithmicMedia or a Google generative claim generator. Then read the certificate independently: valid and trusted is the strong case, valid but untrusted is routine and says nothing bad about the file. XMP fields carrying Google AI strings corroborate the manifest but are unsigned and editable on their own.
Sample status and evidence profile
| Bundled Imagen image sample | Not bundled yet. This page does not claim a reproduced Imagen test until a licensed original can be published alongside the other samples. |
|---|---|
| Google Veo MP4 | A Google generative-AI video sample is bundled with valid Google C2PA. It validates the Google credential path in a video container rather than Imagen image coverage. |
| Fields worth inspecting | digitalSourceType = trainedAlgorithmicMedia, claim_generator_info, Google / Imagen strings, and the provenance chunk inside an original PNG. |
Signals to Check
| C2PA Content Credentials | Strongest when a valid manifest declares trainedAlgorithmicMedia or a Google generative claim generator. Signature and trust state are reported apart. |
|---|---|
| PNG provenance chunk / JUMBF | Where the credential physically lives in an original PNG. Ordinary viewers skip those bytes, which is why the data appears to be missing. |
| XMP / EXIF | Credit, CreatorTool, Software and Google AI strings support the manifest but carry no signature. |
| File name | Platform-style download names are context only and change the moment the file is renamed. |
Why Imagen and Gemini findings look alike
Both sit inside Google's generative stack, so both tend to produce manifests with the same shape: a trainedAlgorithmicMedia source type and a Google claim generator string. What differs is the product surface a file came through and, with it, whether credentials were attached at all and in which format. That is why the practical advice is identical for both — go back to the original download rather than inferring from the picture, and read the claim generator to see which product actually signed it.
A credential that verifies is worth more than one that merely exists
Provenance tooling is often summarised as "does it have C2PA", but that question is too coarse to act on. A manifest can be present and still fail validation because its certificate expired, or verify cleanly while its certificate sits outside the trust list you happen to be using. Those three states — verified, unverifiable, untrusted-but-intact — call for different responses, so they are reported separately here instead of being collapsed into a single badge.
Limits
- Screenshots, chat-app copies and social re-encodes usually remove the C2PA manifest and XMP data entirely.
- Whether credentials are present varies with product entry point, file format and export path.
- No bundled Imagen sample is published yet, so this page describes field coverage rather than a reproduced result.
- No signal found means this copy carries no readable evidence — it is not a conclusion about origin.
Use the C2PA validator for manifest detail; the EXIF/XMP reader covers the supporting fields.
FAQ
Does Google Imagen add C2PA?
Google is among the platforms attaching Content Credentials to generative output, but presence depends on the product entry point, format and export path, and later processing can remove them. Check the original download.
What is the strongest Imagen signal?
A valid manifest declaring digitalSourceType = trainedAlgorithmicMedia, or a Google generative claim generator — materially stronger than an XMP string or a file name.
Why does a valid file read as untrusted?
Valid means the signature verifies and the file is unaltered; trusted means the certificate chains to a root in the list being used. A genuine file can be both at once.
Is my file uploaded?
No. The manifest and metadata are parsed locally in your browser.
Sources and verification records
- AICheck365 public sample records — Which Google samples are bundled, their provenance, test dates and known limits.
- C2PA Technical Specification 2.2 — Manifest structure, assertions and signature verification.
- IPTC Digital Source Type vocabulary — Definitions of controlled values such as trainedAlgorithmicMedia.
- AICheck365 methodology — How claims, signatures and evidence tiers are reported.