AI Detection Tool for Fanfic Built on Fragile Technical Foundation

A browser extension flagging Claude-generated text on Archive of Our Own relies on CSS class markers—implementation details, not durable forensic signals. The tool only works when text is pasted directly from Claude; running output through Google Docs or Word strips the markers. Anthropic could disable detection with any routine interface update. The extension catches procedural laziness, not intentional obfuscation.