I Hate Watermarks is an open-source AI watermark remover that reveals and strips invisible characters from text produced by ChatGPT, Claude, Gemini and other tools. Paste your text to see every hidden character highlighted in place with its name and code point, then copy or download a clean version.
It removes zero-width spaces and joiners, byte-order marks, soft hyphens, direction overrides, variation selectors, Unicode tag characters and exotic spaces, while keeping invisible characters your text needs, such as emoji joiners and RTL marks. Optional settings normalize lookalike spaces and letters or apply NFKC normalization.
Everything runs in your browser, so your text never leaves the page. It is upfront that it can't remove statistical watermarks like SynthID-Text. An agent skill brings the same engine to Claude Code, Codex and other coding agents, and the MIT-licensed TypeScript library has no dependencies.
Key features include highlighting hidden characters with their name and code point, removing zero-width spaces, Unicode tags and exotic spaces, and preserving emoji joiners and RTL marks unless you opt in. Optional lookalike letter replacement and NFKC normalization give you more control over the cleaned output. Because the app runs fully in the browser, your text never leaves the page. There is also an agent skill for Claude Code, Codex, Cursor and more, so the same engine can be used inside modern coding agents.
The tool is built for writers, students, professionals and developers who use AI-generated text and want to detect and remove hidden Unicode characters. Whether you are cleaning up a document before publishing, checking text from an AI assistant, or building a workflow that needs clean Unicode input, I Hate Watermarks shows what is really inside your text and lets you copy a clean version.
It is free and open source under the MIT license. The MIT-licensed TypeScript library has no dependencies, and the project is available on GitHub. If you care about text hygiene, privacy, and open tooling, I Hate Watermarks provides a transparent, browser-based way to inspect and clean invisible characters without sending your content to a server.
The tool is honest about its limits: it can't remove statistical watermarks like SynthID-Text. What it does do is make hidden characters visible and give you control over which ones to keep or remove. You can see every hidden character highlighted in place with its name and code point, then copy or download a clean version. Optional settings normalize lookalike spaces and letters or apply NFKC normalization. It preserves invisible characters your text needs, such as emoji joiners and RTL marks, unless you opt in to removing them.
I Hate Watermarks is a focused utility for anyone who wants to understand and clean the invisible parts of AI-generated text.
