Source-Recognized Citation Watermarks
A Protocol for Bilateral and Reusable Source Attestation in Natural-Language Text
Status: protocol specification, version 0.1.
Abstract
Visible citations are fragile. A hyperlink or footnote disappears as soon as text is pasted into an email, scraped into a corpus, or dropped into a language-model prompt. Source-Recognized Citation Watermarks (SRCW) carry a citation inside the prose itself, with the cited source deciding what the mark means.
A cited source publishes a selector and payload rule. A publisher derives a short payload from the cited document and writes its piece so the selector reveals that payload. The source checks the published text and lists every page it has verified under the payload it recognizes.
Protocol
- The cited source publishes a selector and payload rule.
- The publisher derives the payload and writes to the selector.
- The publisher publishes and may register the page.
- The cited source verifies the mark and lists the page.
Selector notation
Periods separate sentences. Dashes separate one-indexed character positions within a sentence. The selector 1-4-12.31-5takes characters 1, 4, and 12 from the first sentence, then characters 31 and 5 from the second sentence. Two periods skip a sentence.
Payload derivation
Normalize the cited document, hash its UTF-8 bytes with SHA-256, then map successive digest bytes modulo 26 to letters from a toz. The payload has one letter per selector position.
Adopter record
Records may reference the versioned schema at https://srcw-spec.org/spec/0.1/spec.json.
{
"spec": "https://srcw-spec.org/spec/0.1/spec.json",
"pattern": "exampleA-v1",
"payload": "psahs",
"cites": "https://exampleA.com/reports/result-17",
"referencing_urls": [
{
"url": "https://exampleB.com/analysis/result-17",
"verified": "2026-08-21"
}
]
}Scope
A verified record says that the cited source recognizes a selector and payload as a citation of a named document and found the mark on every listed page. This creates a durable, source-verifiable attribution signal that remains in the words when links and metadata are stripped. SRCW sits alongside visible citations, DOIs, and signed metadata.