



The product
AI Watermark Detector analyzes any text for signs of AI generation — with three clearly separated signals, never blended into one misleading score. INVISIBLE CHARACTERS & HOMOGLYPHS 100% verifiable offline: the app looks for invisible Unicode characters and letters from other alphabets substituted for Latin ones — real techniques used to hide manipulated text. AI SCORE (powered by Mistral) A writing-style estimate: probability the text is Claude's, probability it's AI in general, and if not, the most likely model. Clearly labeled as a statistical hint, never proof. OFFICIAL CLAUDE WATERMARK Anthropic confirmed it embeds a real cryptographic watermark in Claude-generated text. As soon as their public detection API ships, the app connects to it automatically — no update needed. FEATURES • Paste text or import a PDF/text file • Share Extension: check text from any app • Block-by-block analysis for long texts • Compare mode: see if an edit changed the signals • Full analysis history, stored only on your device • Available in English and French TRANSPARENCY FIRST This app never claims to prove a text is human or AI-written with absolute certainty. Every signal is labeled with its real reliability level.
LYNXION