Content that moves freely across the internet without losing its origin, identity, or discoverability.
You can compress, resize, rotate, re-upload, change the file name, or strip the meta data. Content Anchoring still works, even on the moon.
Over 30 years of content identity that doesn't travel, defeated
Every system for proving content identity assumes the proof lives somewhere outside the content: in attached metadata, an embedded watermark, a platform database, or a rights registry. Strip the metadata and the identity is gone. Re-encode the file and the watermark breaks. Take the content off-platform and the record doesn't follow. Change a single pixel or byte, and continuity breaks. The content itself has always been passive. It has no idea what it is, where it came from, or what rights it carries.
Content anchoring derives identity from the structural behavior of the content itself. It survives re-encoding, cropping, and platform changes because the identity of the content is the content, not a label on top of it. Meanwhile, a fabricated image has no legitimate lineage. A synthetic video cannot construct origin retroactively. The absence of ownership becomes the proof of forgery. Rights travel with the work, always, without anyone's cooperation.
The next decade of content distribution will require deterministic autonomy - media that can move freely across platforms and still be discoverable. Content anchoring is that architecture.
This isn't theoretical
A working prototype matches and deduplicates images entirely in the browser, in under 800 lines of vanilla JavaScript, with no server, no model, and no inference pipeline. It was tested against a corpus of public domain NASA photographs: images that have been compressed, resized, and re-uploaded across various space news outlets.
Compare that to the two dominant alternatives. Perceptual hashing — pHash, dHash — is fast and simple, but brittle. A moderate crop, a color shift, or a re-encode is often enough to break the match. It was designed for near-identical duplicates, not content that has lived on the internet. Inference-based deduplication — embedding models, vector similarity search — handles variation better, but it requires a trained model, a GPU or inference endpoint, and a vector index that has to be built, maintained, and queried at scale. It is expensive, opaque, and dependent on infrastructure that can go wrong.
Content anchoring requires none of that. Identity is derived structurally, runs client-side, and produces a deterministic result. The same content, regardless of how it has been transformed, resolves to the same identity. That's not a heuristic. It's a structural fact.
One of 16 patent applications. Structural provenance for content that moves.
Patents pending. No guarantee of issuance or scope. No rights granted by this page. Any license requires issued claims (if any) and a separate written agreement.
