Google Lets Users Remove Visible AI Watermarks, Betting on Invisible Provenance That Isn't Yet Deployed Where People Actually See Content

On August 14, 2026, Google announced users can remove the visible Gemini sparkle from AI-generated images and video, leaving invisible SynthID and C2PA metadata intact. The toggle is excluded in the EU and South Korea, where visible watermarks are legally mandated. No SynthID detector runs at consumption points on most social platforms, feeds, or messaging apps. Reuters testing in July 2026 found cropping can defeat SynthID.