AI providers started marking their output: Anthropic watermarks Claude text and signs generated files, OpenAI attaches Content Credentials and SynthID to every ChatGPT image, Google watermarks across four modalities. What has been missing is a place that tells you plainly which of those marks you can actually check, and how — without dressing up guesses as detection.
This site does three narrow things well: deterministic inspection of text for hidden Unicode signals, browser-side validation of C2PA Content Credentials, and a maintained, primary-source-backed map of provider watermark status. Everything runs locally in your browser; the methodology page defines every label we use.
IsWatermarked? is not affiliated with Anthropic, Google, OpenAI, Adobe or the C2PA. Content on provider pages is editorial, written from the primary sources linked on each page. Site editorial is published under the bylineIsWatermarked Research.
The site is operated by “WebSight Group” Ltd, a company registered in Bulgaria (UIC 208723539).
Found an error in a provider fact? The fastest way to get it fixed is to point at a primary source that contradicts us — we treat that seriously.