Claude AI content provenance
Claude watermark checker
Track the current status of Claude’s embedded text watermark and signed C2PA file provenance. Anthropic documents marking for supported models, while third-party detection details are still forthcoming.
- Model-level text watermark
- Signed C2PA file provenance
- Official detection pending
Provider verification
Claude Watermark Status is coming soon
We are connecting and validating the provider-specific verification required for dependable results. This page will open when that check is ready.
Uploads are paused while verification is completed
Use the AI Watermark CheckerWhat the Claude text check reports
Anthropic says supported Claude models embed an imperceptible watermark directly in generated text and attach signed C2PA provenance metadata to supported files. The company has not yet published the third-party detection mechanism, so this checker remains closed until that verification can produce dependable results.
Model-level text mark
Anthropic describes a watermark embedded in generated wording, not a collection of removable zero-width Unicode characters.
Signed file provenance
Supported Claude-generated SVG, PNG, and JPG files can carry digitally signed metadata based on the C2PA standard.
Official detection pending
Anthropic says it will support users and third parties in detecting marks and will publish technical details in forthcoming documentation.
No unsupported verdict
A missing mark cannot prove human authorship, and this page will not turn ordinary Unicode or writing style into Claude attribution.
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.
- Literal Unicode formatting-character inventory with codepoints, counts, positions, and optional cleanup.
- A precise boundary and official path for provider-controlled verification when one exists.
Frequently asked questions
Does Claude watermark generated text?
Yes for supported models. Anthropic says Claude models launched on or after August 2, 2026 embed an imperceptible model-level watermark in generated text, and it is working to add marking support to earlier models.
Can this page detect Claude's official text watermark?
Not yet. Anthropic has announced that it will support third-party detection but says the technical detection details are forthcoming. This page remains in a coming-soon state until a dependable compatible check is available.
Is the Claude watermark made from zero-width Unicode characters?
Anthropic describes the watermark as embedded at the model level and designed to travel with copied text. It does not identify zero-width Unicode as the official watermark. Hidden-character inspection is therefore a different test.
Does Claude add C2PA provenance metadata to files?
Anthropic says supported Claude-generated files such as SVG, PNG, and JPG can carry digitally signed provenance metadata based on the C2PA standard. Support can vary by model, product, platform, and file type.
Does a missing Claude watermark prove text is human-written?
No. Anthropic notes that marks may be absent or undetectable in older-model output, short passages, heavily edited or translated text, mixed writing, or unsupported products and file types.