Google is making the visible watermark on AI-generated media optional. A new toggle, announced by Gemini vice president Josh Woodward, covers output from the Nano Banana, Omni and Lyria models in Gemini and the Flow video editor, with Search support on the way, and rolls out in the coming days under Settings, Media Watermark.
Invisible markers remain on every file. Woodward said invisible SynthID watermarks and C2PA metadata still tag each generation, so Gemini and Search can keep identifying AI-made content even when the visible mark is switched off, a balance he described as creative control against safety.
Google also open sourced Credentio, a C++ library for local C2PA validation that the company says already powers dozens of its provenance-compliant products across tens of billions of generated assets. Developers can embed it to check content credentials inside their own apps without shipping files to cloud services.
The timing is notable: Anthropic only recently began embedding marks in Claude’s text and files to meet EU requirements, while Google’s toggle moves the opposite way on visibility. Either way, the provenance layer underneath stays on, keeping the AI origin of media verifiable after the visible stamp disappears.
