Claude Watermark Explained: What Comes Off and What Doesn't 👀 Since 11 August 2026, Claude adds an invisible watermark to everything it makes — text, images, code — across Claude, Claude Code, Cowork and the API. No opt-out. It's the EU AI Act. There are two completely different marks, and everyone's mixing them up. 📁 FILES (.png .jpg .svg) A signed C2PA content credential sitting in the metadata. Any free AI watermark remover can strip this in seconds. 📝 TEXT — SynthID-Text Not invisible Unicode. Not em-dashes. Not metadata. It's which words Claude picked. There's nothing hidden in the text to simply delete. Anthropic's own explanation: light editing won't remove it. A complete rewrite where every word is replaced will. The open-source project watermarks-remover has 15k+ stars in two weeks. It handles the file side by removing the metadata and handles the text side by rewriting it — although its own README admits this can flatten your tone. And here's the big catch: Nobody can verify the text side yet because Anthropic hasn't released the detector. Comment WATERMARK and I'll send you both links 👇 #Claude #ClaudeAI #AIWatermark #Anthropic #aitools how to remove the claude watermark / what is claude's ai watermark / claude synthid text explained / free ai watermark remover / c2pa metadata remover / does claude watermark text - @theaiimpact