To a crawler: usable or not, and on what terms. To you: a timestamped claim if your work is used. To anyone reading an AI answer: where the knowledge came from.
Most work enters AI systems with no record of what was taken, whose it was, or on what terms. Registration ends that for yours: usable or not, your call, machine-readable. If your work is used, you have a claim.
Check the fingerprint and you know what you may use, where it came from, and on what terms, with a signed receipt for the regulator. What Europe requires, answered per work.
A claim you cannot source is a claim you cannot act on. When the work behind an answer is registered, the source exists: a named owner, on the record.
The file never leaves your computer.
Your browser scans the file where it sits and sends only the fingerprint, the number that identifies the work in any exact copy. A work can carry several fingerprints, one per format it ships in. You choose the terms the record carries, from the conditions on the register page. Free. The email you register with is your account: sign in with it to see your records and which AI crawlers have read your pages. A revision is a new registration; the chain stays intact.
Written work today: documents, publications, reports, specifications, code, and datasets that move as exact copies. Photographs, video, and audio are on the roadmap; until their fingerprint ships, we do not claim them.
A fingerprint alone is a number you hold by yourself. A record is countersigned, timestamped, published, and answered for.
That is what turns an assertion into evidence. A superseded record is marked, never deleted. History stays readable, permanently.
You receive the record as a JSON export, a public verify URL, and a page embed you can place anywhere.
A record, as anyone sees it at its public URL. Details invented for illustration.
If you control a website, make the registration impossible to miss.
A crawler should see the answer before it has to go looking. One line in your page, one file at your domain root, and the public record URL for takedowns, disputes, and licensing conversations.
<!-- in the page, where a machine reads first --> <script type="application/ld+json" class="alltio-record"> { "record": "https://alltio.com/r/9K42-D317", … } </script> # once, at your domain root, covering everything you publish /.well-known/alltio.json
An ingestion pipeline computes the same fingerprint and asks one question: who owns this work, and on what terms?
The answer tells a lab what it may use, where it came from, and on what terms, with a signed receipt as proof for the regulator. Since 2 August 2026, providers of general-purpose models must demonstrate exactly that under the EU AI Act. Checked use is defensible. Unchecked use is what ends up in court.
Record verification is live today. Fingerprint lookup at ingestion scale is a published draft specification.
$ curl https://alltio.com/api/verify/9K42-D317 { "registered": true, "terms": "attribution", "filed": "2026-08-02T14:31:09Z", "countersignature": "valid" }
Alltio keeps the count, never the funds.
Built to clear the deals both sides already make: a standard contract, signed usage reports, settlement on partner payment rails, and a dispute path with an audit trail. Nothing is charged today.
That is the design. Registration gives you a timestamped, verifiable claim you can use today in provenance disputes, takedowns, and licensing conversations that have nothing to do with us. Everything above is what that record plugs into as the clearing layer comes online. Register because the record stands on its own; the rest is what it was built for.
Create your record