Content Provenance & C2PA Checklist
If you publish AI-generated or AI-assisted content, provenance is how a viewer can later check that it is what it claims to be. Work through the five layers — a C2PA / Content Credentials manifest, a cryptographic signature, a durable watermark, clear disclosure, and an unbroken chain-of-custody — and score your coverage. Watermarks alone can be stripped, so these layers back each other up. And provenance is a dial, not a switch: the full stack is for high-trust content, not every internal draft.
Full stack justified — contested or consequential authenticity. Target ~90/100.
C2PA manifest
Cryptographic signing
Watermarking
Disclosure & labeling
Chain-of-custody
0/100
Weak
Marker shows the target for this content type.
For high-trust (news, evidence, legal, finance), this coverage is thin. The provenance you do have may not survive a determined challenge. Close the biggest gaps below.
Biggest gaps
- 1.[C2PA manifest] A C2PA manifest records how the asset was generated and edited.
- 2.[Cryptographic signing] The manifest is signed by a key tied to a trusted, verifiable signer.
- 3.[Watermarking] A durable (hard-binding) watermark is applied, not only metadata.
- 4.[Disclosure & labeling] Content is clearly disclosed as AI-generated or AI-assisted in human-readable form.
What this checklist is
- A scored checklist for adding real provenance to AI-generated or AI-assisted media and text, across five layers: C2PA / Content Credentials manifests, cryptographic signing, durable watermarking, disclosure and labeling, and chain-of-custody from capture to publish.
- C2PA is the open standard from the Coalition for Content Provenance and Authenticity; Content Credentials is its consumer-facing label. A C2PA manifest is cryptographically signed metadata about how an asset was made and edited.
- Watermarks are not tamper-proof: soft-binding watermarks (metadata, invisible markers) can be stripped or degraded by re-encoding, screenshots and format conversion. Durable watermarking helps but is not a guarantee — layer it, do not rely on it alone.
- Provenance is a dial, not a switch: full C2PA signing and chain-of-custody are worth it for news, evidence, legal and other high-trust content; they are overkill for ephemeral internal drafts and low-stakes assets where simple disclosure is enough.
- Fully client-side and free. Your selections never leave the browser; export the checklist as text.
Why this matters
As generated media becomes indistinguishable from captured media, the question shifts from “is this real?” to “can I verify where this came from?” Content provenance is the infrastructure for that answer. Done well, it lets a newsroom prove a photo is authentic, lets a platform label synthetic media honestly, and lets a court trust an exhibit. Done naively — relying on a single watermark that a screenshot removes — it offers assurance that evaporates the moment it is tested, which is worse than offering none.
The discipline is to layer the controls and to apply them in proportion to stakes. Signed C2PA manifests plus durable watermarks plus disclosure, carried through an unbroken chain-of-custody, make provenance resilient to any single layer being stripped. Reserve that full stack for the content where authenticity is contested or consequential, and keep low-stakes content to simple disclosure — so the effort lands where it protects something real.
Frequently asked questions
What is C2PA and Content Credentials?
C2PA stands for the Coalition for Content Provenance and Authenticity, an open technical standard for recording the provenance of digital content. A C2PA manifest is a block of cryptographically signed metadata attached to an asset that records how it was created and edited — for example, that it was generated by a particular AI model and then cropped. Content Credentials is the consumer-facing name and the little "CR" marker used to display that information. Together they let a viewer, with the right tools, check who made a piece of content and what was done to it.
Are watermarks enough to prove something is AI-generated?
No. Watermarks are useful but not tamper-proof. Soft-binding approaches — provenance stored in metadata, or invisible statistical markers in pixels or text — can be removed or degraded by re-encoding, screenshotting, cropping, or converting formats, and some can be deliberately stripped. Durable (hard-binding) watermarks are more robust but still not a guarantee. The reliable posture is defense in depth: combine a durable watermark with a signed C2PA manifest and clear disclosure, so that if one layer is stripped the others still carry the signal.
What does this checklist score?
It covers five provenance layers for a piece of content: a C2PA / Content Credentials manifest, a cryptographic signature from a trusted signer, a durable watermark, clear human-readable disclosure and labeling, and an unbroken chain-of-custody from capture or generation through every edit to publication. For each item you mark whether it is in place, and the checklist gives you a coverage score plus the gaps that matter most for your content type.
When is full content provenance worth it, and when is it overkill?
Provenance effort should match the stakes and trust requirements of the content. News and journalism, legal and evidentiary material, official communications, financial disclosures and anything likely to be contested or impersonated justify the full stack: signed C2PA manifests, durable watermarks and chain-of-custody. Ephemeral internal drafts, low-stakes marketing variants and throwaway assets do not — there, a simple disclosure that content is AI-assisted is proportionate, and signing everything just adds cost and a false sense of rigor. The checklist flags when you are likely over-investing.
This tool is free and always will be. Would a Pro workspace — saved designs, team sharing and exports — be worth paying for?
Show me Pro →No account needed to use this tool — this just tells us whether Pro is worth building out.Educational — general guidance, not security or legal advice.