The era of unprovable AI writing just ended. Starting August 2, every new Claude model weaves an invisible watermark into the words it generates. Copy it, paste it into an email, a blog post, a college essay, and the watermark travels with the text. Anthropic will publish the detection method so anyone can check, which means your professor, your boss, and the platform you post on can all run the same test you can. Here's the part almost nobody knows. Google has been doing this to Gemini since…