How Blockchain Timestamping Works

Services like WordProof create a cryptographic hash of your content and record that hash on a blockchain - typically a public blockchain like Ethereum or a purpose-built chain. The hash is deterministic: the same content always produces the same hash, and any change produces a different hash.

To verify a piece of content against the blockchain record, a verifier takes the current content, computes the hash, and checks whether that hash appears in the blockchain at the claimed timestamp. If it does, the content existed in that form at that time. If it does not, the content was either never recorded or was modified after recording.

The record is real and the cryptographic guarantee is sound. The problem is not with the cryptography. The problem is with the architecture: the record and the content are separate. When the content moves, the record stays behind.

The Separation Problem

A publisher timestamps an article on-chain and distributes it to AP. AP distributes to 1,500 subscriber outlets. Those outlets publish on their own CMS, which strips metadata. The article is scraped by aggregators, some of which strip headers. Eventually, a version ends up in an AI training corpus.

At every step in that chain, the blockchain record exists on the blockchain. But the content that traveled through the chain has no connection to that record. There is nothing in the article text or the file that points to the blockchain transaction. Verification requires knowing where to look and having the original hash to compare against.

The AI company that received the content in its training corpus has no way to discover the blockchain record. They would need to know the original publisher, look up the blockchain registration, obtain the transaction details, and perform the hash comparison. In practice, this does not happen. The ownership record is effectively invisible to anyone downstream.

How Embedded Provenance Solves This

C2PA manifests are embedded in the content itself - in the file container for images and media, and in invisible Unicode markers for text. The proof does not live on a separate server or blockchain. It is part of the content.

When that content moves through distribution channels, the proof moves with it. When AP distributes the article, the manifest is in the text. When an aggregator scrapes it, the manifest is still in the text. When an AI company ingests it into a training corpus, the manifest is still in the text.

Any party downstream - without needing to contact the publisher, without needing to access a blockchain, without needing any external lookup - can verify the content's provenance using open-source C2PA verification libraries. The proof is self-contained and self-verifying.

Technical Architecture Comparison

The Rights Terms Advantage

Blockchain timestamping records that content existed. It does not record what rights apply to it. A blockchain timestamp does not make content's licensing terms machine-readable. It does not encode Bronze/Silver/Gold tier permissions. It does not create the formal notice that converts innocent infringement to willful infringement.

C2PA manifests include structured rights assertions. The manifest can encode whether content is available for indexing, RAG use, or training. It can specify attribution requirements, geographic restrictions, and commercial versus non-commercial use permissions. These terms are machine-readable - AI systems can parse them without human interpretation.

When an AI company ingests content with embedded rights terms, they have formal notice of those terms regardless of whether they read them. The embedded notice is the mechanism that shifts the legal burden from innocent to willful infringement. A blockchain timestamp without rights terms does not create that shift.

When Blockchain Timestamps Have Value

Blockchain timestamping is useful for priority disputes - proving you created something before a competitor did. Patent applicants, inventors, and researchers use blockchain timestamps to establish a dated public record that does not depend on a trusted third party.

In this narrow use case, the separation of record and content is not a problem. The goal is a dated public record, not a proof that travels with the content. A blockchain timestamp serves that goal.

For content that will be distributed, syndicated, and consumed by third parties who need to discover its provenance - published articles, licensed images, audio recordings, video content - embedded provenance is more useful than a blockchain record that requires a separate lookup chain.