Browser tools claiming to strip AI watermarks from Claude, ChatGPT, and Gemini outputs mostly clean up Unicode characters and reformat punctuation, according to the article. These tools target zero-width characters and invisible Unicode markers, with some claiming to regenerate text through a different model to defeat "statistical watermarks." At least one tool names Claude 3 Opus, Sonnet, Haiku, and Claude 2 as supported targets. Removing zero-width characters is technically straightforward, but the stronger claim of breaking a statistical watermark is unverified by any independent source. One research finding warns that certain Claude watermark mechanisms won't appear on basic character scans, meaning tools might scrub formatting noise and miss the actual signal. The tools also normalize em dashes, convert smart quotes, and collapse blank lines, suggesting text formatting is the real use case. According to a TechCrunch report, Google's Gemini 2.0 Flash can remove watermarks from images, while Claude called watermark removal "unethical and potentially illegal" when tested on the same task. No independent source has verified that any browser tool reliably defeats a genuine watermark under real conditions. The article concludes that users who found these tools on an X thread likely received a Unicode scrub and tidier punctuation rather than a clean break from AI attribution.