Google is loosening the most obvious label on Gemini-made media without abandoning its effort to make AI content identifiable. The change hands creators more control, while asking invisible technology to carry the burden of trust.

On Friday, Google said users will be able to switch off the visible watermark applied to output from its generative AI tools, including images, videos and songs. The option is set to cover the Nano Banana, Omni and Lyria models, and will be available in Gemini and Google’s Flow video editor; Search support is expected to follow. Users will find the control under Settings > Media Watermark as the feature rolls out in the coming days.¹

The practical appeal is clear: a prominent label can limit the usefulness of an image or clip in professional and creative work. Google’s answer is to make that label optional rather than treat it as the sole safeguard. The company says disabling the visible mark will not remove its invisible SynthID watermark or C2PA-related metadata, tools designed to preserve a record that media was generated or altered by AI.¹

Josh Woodward, Google’s vice-president for Gemini, framed the shift as a compromise between user flexibility and disclosure. “We’re striking a balance here between creative control and safety,” he said, adding that users should still be able to use Gemini or Search to determine whether an image was AI-generated.¹

Google is also open-sourcing Credentio, a library intended to help developers build local validation into their own apps. The move lands amid a wider fight over how conspicuously AI output should be labelled, following criticism of Anthropic’s watermarking of Claude-generated text and files to meet EU rules.¹