Claude Watermark Checker & Detector | Unicode Scan
Check whether Anthropic's official Claude watermark detector/API is available, then scan text locally for hidden Unicode. The tool is private and free, but it does not produce an official watermark verdict.
Your text is analyzed locally in your browser for the currently available checks.
0 characters · 0 words
Analysis results
How to read the result
Inspect text integrity without overstating provenance
Start with the official verification status. When it is unavailable, the rest of the report is a local integrity inspection: it can reveal hidden Unicode and explain sample structure, but it cannot identify a human or AI author.
Paste a representative passage and run the local analysis.
Review every unusual character in context before removing it; some are legitimate language or formatting characters.
Use the cleaned-copy preview only when you have confirmed that the selected invisible characters are unwanted.
WordPress publishing
Need a clearer review and publishing process?
I can help implement WordPress editorial roles, review steps and custom publishing features. Describe the problem and your website; we can agree a practical scope. This checker does not prove who wrote a passage.
Clear boundaries for a privacy-first local inspection.
Can this tool verify an official Claude watermark?
Not today. Anthropic has announced a Claude watermark detection API, but it is not yet publicly available. This tool therefore reports official verification as unavailable instead of inferring authorship from hidden characters, punctuation, or writing style.
Does Claude currently watermark generated text?
Anthropic says Claude models launched in the EU on or after August 2, 2026 support machine-readable marking at launch, while support for earlier models is being added progressively. A missing or unavailable verification result therefore cannot prove that text is human-written.
Is the Claude watermark hidden Unicode or SynthID?
Anthropic says Claude uses a version of Google DeepMind's SynthID-Text approach. Nothing is added as a secret character: the statistical signal is created through otherwise-valid word choices during generation. Hidden Unicode is a separate text-integrity issue.
How can I detect Claude's official text watermark today?
You need Anthropic's compatible detector and watermark key. Anthropic says a detection API is coming soon, but no public production API is available for this checker yet. Until that changes, third-party Unicode scans and generic AI scores are not official verification.
Can this tool remove Claude's statistical watermark?
No. The cleaned-copy option removes only selected invisible Unicode characters. Anthropic says Claude's watermark is a statistical pattern in generated word choices, so Unicode cleanup must not be described as watermark removal.
Does hidden Unicode prove that text came from Claude?
No. Zero-width characters, non-breaking spaces, bidirectional controls, and variation selectors can come from copy-paste workflows, editors, languages, accessibility tools, or malicious manipulation. They are inspection signals, not proof of source or authorship.
Does my pasted text leave my browser?
No. The scanner, counts, suitability heuristic, cleaning preview, and copy action run in your browser. This page has no text-analysis API endpoint and does not store pasted content.
What does sample suitability mean?
It is a transparent structural heuristic based on length, prose continuity, repetition, and code-like content. It only indicates whether text could be useful to inspect if an official statistical method becomes available; it is not a watermark result.
Continue with the AEO Toolkit
Use the local text inspection alongside page-readiness and structured-data checks.