Evidence, not guesses.
AIWaterMarkCheck was built to provide honest, verifiable clarity in an increasingly synthetic digital landscape. We prioritize cryptographic proof, clear limitations, and complete user privacy.
Open Standards First
We inspect cryptographically signed C2PA Content Credentials and open reference implementations rather than relying solely on opaque, uncalibrated AI classifiers.
100% In-Memory Privacy
No user uploads, analyzed texts, or scraped URLs are ever stored on disk or in a database. Submissions are processed transiently in RAM and discarded immediately.
Calibrated Transparency
We never claim a missing signal proves human authorship. When evidence is incomplete or inconclusive, we report the uncertainty clearly and honestly.
Why we created AIWaterMarkCheck
The internet is experiencing an unprecedented surge in synthetic media. At the same time, many commercial AI detectors market unrealistic 99% accuracy rates that frequently penalize non-native English speakers, creative prose, and legitimately edited photos.
We believe that content consumers and creators deserve a tool that clearly distinguishes between verifiable cryptographic provenance (such as C2PA manifests signed by Adobe or OpenAI), metadata assertions, and statistical model signals.
Core Commitments
- Distinction between evidence types: A signed C2PA certificate is treated as cryptographic fact; a neural network score is presented as a probabilistic signal.
- No paywalls for safety: Foundational watermarking and C2PA inspections should be accessible to everyone—journalists, educators, fact-checkers, and everyday internet users.
- Zero corporate conflicts: AIWaterMarkCheck is an independent project. We are not owned by or affiliated with Google, OpenAI, Anthropic, Adobe, or Meta.
Responsible Interpretation
Watermark detection is a fast-evolving discipline. Compression, screenshots, re-encoding, and intentional adversarial modifications can easily strip or blur watermarks. As a result, our reports should be used as supporting evidence in investigative workflows, not as autonomous verdicts for high-stakes decisions.
Inspect an image, text passage, or website
Experience multi-layered detection with complete privacy and zero data retention.