Claude Watermark is a browser-based tool that checks text for Claude- and ChatGPT-related artifacts, cleans verifiable traces, and helps resample prose.
Claude Watermark is a browser-based tool for checking text for Claude- and ChatGPT-related artifacts and cleaning the ones it can verify. It focuses on concrete traces such as HTML markup, invisible Unicode characters, non-breaking spaces, and punctuation patterns that can survive copy and paste.
The site also explains Anthropic’s newer statistical watermark for Claude models and is explicit about a key limitation: there is no public detector for that watermark yet. In practice, the product combines a local checker with a paid rewrite flow, so readers can inspect text, remove detectable artifacts, and resample prose when they need to reduce AI-like signals.
Paste text into the checker and it reports concrete artifacts as they appear, rather than assigning a probabilistic score. Findings include HTML class names from Claude’s web interface, invisible Unicode characters, non-breaking and exotic spaces, curly quotes, and em dashes.
The site separates byte-level findings from stylistic signals. Deterministic artifacts are presented with counts and positions, while style-based signals are treated as context rather than proof.
A one-click cleanup path removes the detectable artifacts the site can actually verify, including Claude-related markup and invisible characters. The homepage also describes a rewrite pass that reworks sentence rhythm and length.
The service states that the rewrite pass runs every sentence through a different model to help break the distributional pattern associated with AI-written prose. It is positioned as the only removal method the site says can address Anthropic’s watermarking approach.
Checking runs locally in the browser with no account and no upload. The site also says it is free and unlimited for checking.
Pricing is separated from checking: rewrites use credits, with three free credits on signup and no card required. The pricing page also says failed jobs are refunded automatically and credits never expire.
Useful when text was copied out of Claude and you want to confirm whether markup, zero-width characters, or other copy-paste artifacts came along with it before posting or sharing.
Helpful for writers who want to remove detectable artifacts from a draft and then rewrite the passage so it reads more naturally after cleanup.
Useful for forum, fandom, and CMS workflows where pasted HTML classes or invisible characters can trigger moderation, display problems, or community-specific detectors.
Helpful when you need a clear distinction between what is actually present in the bytes and what only looks stylistic, so you can decide whether cleanup is warranted.
Useful when you want a private check before sending text onward, since the checker runs locally in the browser and does not require an account.
Checking runs in the browser, so the text is not uploaded and no account is required. Cleaning and rewriting use credits, with 3 free credits on signup and no card required.
The checker can verify concrete artifacts in the text, including Claude-related HTML class names, zero-width or invisible Unicode characters, non-breaking and exotic spaces, and stylistic signals such as curly quotes and em dashes. It cannot verify Anthropic’s statistical watermark itself because no public detection key or released detector exists yet.
No. The site says Anthropic’s watermark is carried by word choice, so deleting invisible characters does not remove it. The removal method the site relies on is rewriting or resampling the text with a different model.
The site says checking is instant as you type. Rewriting usually takes a few seconds for a typical page, and the time scales with document length.
The site says the rewrite pass is processed by its model provider and not stored afterward. It also says checking stays in your browser, so there is nothing to upload for that step.
Writecream AI Content Detector & Checker is a free tool to check whether pasted text appears written by AI or a human, with an AI content percentage.
Winston AI is an AI content detector and plagiarism checker for text and images, helping educators, publishers, and institutions review originality and share results.
Verol is a Chrome extension that detects AI hallucinations and verifies claims in real time in ChatGPT, Claude, Gemini, and Grok for fast browser-based fact checking.
Honestly turns social posts into structured product signals for brands. Get sentiment analysis, trend tracking, affiliate discovery, and social proof.
tata.run is a professional free AI content detector that provides a one-click 'reduce AI' feature to significantly lower AI detection rates.
学术猹 by NetEase Youdao offers paper checking, AI detection, rewriting and AI reduction for Word, PDF and TXT files, with a free daily check.
