Skip to main content
v2026.1714 entries · CC-BY 4.0
Dictionary termTrack AStablev2026.2

Watermarking (AI output)

The embedding of a statistical, cryptographic, or visible signal into AI-generated content at the time of generation, allowing later identification of that content as machine-produced — distinct from post-hoc detection which infers AI origin from features of the output alone.

ByCASRAI Editorial Board
· Last updated 21 May 2026

Examples

Worked examples

  • Is an instance

    Google SynthID embedded in Imagen outputs

  • Is an instance

    Statistical token-selection watermark in some LLM APIs

Counter-examples

Looks similar, but isn't

  • Not an instance

    A visible 'Made with AI' caption added by a user is a disclosure but not a watermark in the technical sense

Editorial commentary

Watermarks may be visible (a corner badge) or invisible (statistical bias in token selection, frequency-domain signal in image pixels, C2PA-style cryptographic provenance metadata). For scholarly governance, watermarking is one mechanism for AI provenance but is not yet universal across generators and is sometimes removable.

References

  • Kirchenbauer et al. 2023 ‘A Watermark for Large Language Models’ ICML
  • C2PA Coalition for Content Provenance and Authenticity Specification (2023)

Also known as

AI watermarking · Generative watermark

Machine-readable encodings

Use in your systems

JATS XML <role> element
xml
<role vocab="credit"
      vocab-identifier="https://casrai.org/dictionary/"
      vocab-term="Watermarking (AI output)"
      vocab-term-identifier="https://casrai.org/dictionary/term/watermarking-ai-output" />
Schema.org DefinedTerm (JSON-LD)
json
{
  "@context": "https://schema.org",
  "@type": "DefinedTerm",
  "name": "Watermarking (AI output)",
  "identifier": "https://casrai.org/dictionary/term/watermarking-ai-output",
  "description": "The embedding of a statistical, cryptographic, or visible signal into AI-generated content at the time of generation, allowing later identification of that content as machine-produced — distinct from post-hoc detection which infers AI origin from features of the output alone.",
  "inDefinedTermSet": "https://casrai.org/dictionary/domain/generative-ai-use-and-disclosure/",
  "url": "https://casrai.org/dictionary/term/watermarking-ai-output",
  "sameAs": [
    "AI watermarking",
    "Generative watermark"
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}

Adopted by research universities worldwide

University of Cambridge logoColumbia University logoUniversity of Edinburgh logoHarvard University logoMassachusetts Institute of Technology logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoUniversity of Cambridge logoColumbia University logoUniversity of Edinburgh logoHarvard University logoMassachusetts Institute of Technology logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logo
  • University of Cambridge logo
  • Columbia University logo
  • University of Edinburgh logo
  • Harvard University logo
  • Massachusetts Institute of Technology logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo

View CASRAI adoption →