Dozens of Claude subscribers on X criticized the decision, expressing concerns that content generated or revised with Claude could still be identified as AI-assisted even after the text is reworked, reformatted, or incorporated into another piece of writing.
Some critics described the policy as “a conspiracy against innocent Claude users.” Others characterized the watermark as a modern-day “scarlet letter,” a mark that could remain identifiable long after any piece of AI-assisted text leaves the platform.
“Who will get caught? You,” a user named visionode wrote in “The student who used Claude to reorganize a paragraph. The journalist who asked the AI to summarize a two-hundred-page transcript. The writer who had creative block and asked for synonyms. Those guys come out of the process with a digital tattoo on their forehead.”
Anthropic disputes that characterization. The company said the watermark “carries no identifying information and can’t be traced to a specific person, organization, or chat.”
The new watermark, based on Google DeepMind’s SynthID-Text technology, allows AI-generated text to be identified without a visible label. Instead, it embeds a statistical pattern into ordinary word choices made by the model. The pattern is invisible to readers, Anthropic said, but can be picked up by someone using a verification key to determine “the likelihood that Claude was involved in writing the text.”
The company said editing can affect whether the watermark can be detected, but acknowledged that minor revisions are unlikely to delete it. “Light editing probably won’t remove the watermark completely,” Anthropic said. “A complete rewrite where every word is replaced will.”