Trust
Verification: checking a record without an account
A provenance record is only worth something if somebody who does not trust you can check it. The strongest check Wakemark offers depends on nothing from Wakemark at all.
The signal ladder
Not every check establishes the same thing, and collapsing them into one green tick would be the easiest way to make the whole exercise worthless. In descending order of strength:
- 1. Hash the file yourself
- Compute the SHA-256 of the file in your hand and compare it to the digest on the record. This needs no account, no API call and no trust in Wakemark — the arithmetic is the same everywhere. It establishes that the bytes are identical to the ones recorded at write time.
- 2. Validate the embedded C2PA manifest
- Where the generating provider embedded Content Credentials, they can be validated against the signer's certificate independently of Wakemark, using any C2PA-capable tool. This is the strongest claim about who produced the file, and it comes from the provider, not from us.
- 3. Look the record up
- Given an asset id, the record itself — digest, provider, model, timestamps, C2PA state, and the provenance chain — can be read back and compared. This one does depend on Wakemark being honest about its own ledger, which is why it sits below the two checks that do not.
What a public lookup returns
A public record is redacted server-side to non-sensitive provenance fields. It carries no prompt, no signed URLs, no storage keys, and no account details, and an unknown id returns the same uniform response as one that has been purged — a stale id is indistinguishable from a never-existent one, so the endpoint cannot be used to enumerate what exists.
Verification after a purge
The digest outlives the bytes. When retention deletes an asset, the tombstone keeps its id, its SHA-256 and its size, so a file somebody downloaded a year ago can still be checked against the record long after Wakemark stopped holding a copy. That is the useful half of a tombstone: what was deleted is the content, not the ability to verify it.
What verification cannot do
- It cannot tell you anything about a file that never passed through Wakemark. There is no record to check.
- It cannot prove a file is AI-generated in general. It can show that a specific generation request produced these exact bytes.
- It cannot resurrect a mark the provider never embedded. Where no manifest exists, the honest answer is
none— and where nothing inspected the file, the honest answer isnot_inspected, which is not the same thing.
Check this yourself
Nothing on this page should have to be taken on our word. Each of these settles part of it, and none of them requires trusting the sentence above it.
- Verify an asset — rung three, run it now: look a record up with no account and see exactly which fields come back.
- c2pa-rs and the C2PA command-line tool ↗ — rung two, independent of us: validate an embedded manifest against the signer's certificate with somebody else's software.
- Custody — rung one, what the digest does and does not prove once your hash matches.