Universal provenance workstation
AI watermark checker: scan C2PA & metadata
Scan images, video, audio, and PDFs for C2PA Content Credentials, declared AI metadata, generator tags, and exact file-level inspection coverage. Pasted-text inspection is coming soon.
- Images, video, audio, and PDF
- C2PA, EXIF, XMP, and IPTC
- Exact inspection coverage
- Checking service
- Add content
- Inspect content
- View report
Step 1 of 4: Checking service.
Service check
Preparing the provenance inspector
Confirming the inspection service is ready before anything is sent.
One file, four provenance layers
Each layer answers a different question. The report keeps signed credentials, editable metadata, embedded provider signals, and visible or textual marks separate so you can evaluate the file precisely.
C2PA Content Credentials
Parsed provenance manifests, when present. Validation, trust, issuer, and generator fields are shown only when the validator returns them.
EXIF, XMP & IPTC metadata
Camera fields, software tags, and IPTC digital source types that may record how a file describes itself. They are supporting evidence, not origin proof.
Provider verification path
Proprietary signals inside pixels, waveforms, or model output are listed separately with the compatible provider check they require.
Visible mark review
Visible logos and overlays are identified as a separate manual-review layer rather than being presented as an automated provider-watermark result.
Evidence you can use
Every report separates cryptographic credentials, declared metadata, and provider-controlled verification. You can see exactly what ran before deciding what the file proves.
- C2PA Content Credentials parsing and cryptographic validation for supported files.
- Declared EXIF, XMP, IPTC, container, and generator metadata evidence.
- A precise boundary and official path for provider-controlled verification when one exists.
Frequently asked questions
What can this scanner actually detect?
It inspects supported file-level provenance signals: C2PA Content Credential manifests, EXIF, XMP, and IPTC metadata, and generator software tags. Results are reported per signal with a clear status. Pasted-text inspection is coming soon.
How are invisible embedded watermarks like SynthID checked?
The report lists proprietary pixel, waveform, and model-level signals under a separate provider-verification step. Use a compatible official provider verifier when one exists; this file scan never substitutes editable metadata for an embedded-signal result. Anthropic does not currently publish a Claude text-watermark verifier.
What does trainedAlgorithmicMedia mean?
trainedAlgorithmicMedia is an IPTC digital-source declaration for content created with a generative AI model. It is useful supporting metadata, but it is editable and therefore is not an independent AI-origin detector result.
Does removing metadata remove AI provenance?
Not necessarily. C2PA metadata and EXIF fields can be removed, but embedded watermarks live inside the pixel, audio, or text statistics themselves and may survive metadata cleaning. This scanner shows which layers a result came from.
Is my file uploaded anywhere?
The file is sent to AICleanify's scanner service, processed on temporary local disk, and deleted after the request. AICleanify does not retain the file; only an anonymous aggregate usage statistic is recorded.