You cannot use a hidden-character cleaner to verify removal of Claude's announced statistical text watermark. That signal is based on patterns in generated word choices. You can, however, inspect and safely clean observable Unicode formatting characters such as zero-width spaces, soft hyphens, unusual spaces, and directional controls. The tool below changes formatting characters only—it does not rewrite your wording, remove C2PA file credentials, or make text “detector-proof.”
Run the evidence check first. Safe clean appears only when actual formatting characters are found.
We will list exact Unicode code points, positions, and readable context.
This cleaner does not detect or remove a word-choice watermark.
You cannot verify removal of Claude's statistical watermark with this hidden-character cleaner. You can inspect and remove observable Unicode formatting characters without rewriting the text.
Copying may strip some formatting or file metadata, but it does not prove that a statistical word-pattern watermark is gone.
No. Anthropic's announced text watermark is statistical. Hidden Unicode characters are a separate formatting issue that this tool can show precisely.
No. WatermarkScope verifies C2PA credentials but does not provide a tool to strip signed provenance information.