Google is making its visible AI watermark optional. On Friday, the company announced that users will be able to turn off the visible watermark on images, videos, and songs generated with its AI tools — while keeping the invisible identification systems that let anyone verify a file's origin.
Josh Woodward, Google's vice president for Gemini, said the toggle will be available for creations made with the Nano Banana, Omni, and Lyria models. The setting can be switched in the Gemini app and in Flow, Google's video editor, with support for Google Search coming later. "We're striking a balance here between creative control and safety," Woodward wrote: the visible mark becomes optional, but invisible SynthID watermarks and C2PA metadata remain, so users can still check via Gemini or Search whether an image was AI-generated. The feature rolls out in the coming days under Settings > Media Watermark.
Google is also open-sourcing a library called Credentio that lets developers embed local validation of AI provenance directly in their apps.
The change marks a shift in the industry's approach to labeling synthetic media: instead of making the label obvious at a glance, the burden moves to invisible, machine-readable provenance — a direction that matters as the EU's AI Act transparency obligations take effect and regulators in California and elsewhere tighten their rules. It also contrasts with Anthropic's recent, more controversial decision to embed watermarks in text and files generated by Claude. Removing the visible mark does not remove the underlying identification data.
