Google is loosening the most obvious label on its AI creations while insisting the underlying trail will remain intact. The move puts creative flexibility and public transparency on a collision course—and asks users to trust invisible safeguards.
On Friday, Google said Gemini users would be able to switch off visible watermarks on AI-generated images, videos and songs. The option is planned for the Nano Banana, Omni and Lyria models in Gemini and Google’s Flow video editor, with Search support to follow. Users will find the control under Settings > Media Watermark as the feature rolls out over the coming days.1
The practical appeal is clear: a conspicuous watermark can make generated material awkward for professional or creative work. A second report captured the change in plain terms: “You can now turn off Google Gemini’s visible watermarks.”2 But Google is not abandoning labeling altogether. It is shifting the responsibility from a mark that viewers can immediately see to technical signals that require a tool to inspect.
Josh Woodward, Google’s vice president for Gemini, framed that distinction as a compromise: “We’re striking a balance here between creative control and safety.”1 Google says the setting will not remove its invisible SynthID watermark or C2PA-related metadata, meaning Gemini or Search should still be able to identify whether an image was AI-generated.1
The company is also open-sourcing Credentio, a library intended to let developers build local validation into their own apps. The announcement follows broader fights over AI disclosure, including Anthropic’s debated watermarking of Claude-generated text and files to meet EU rules. Google’s answer is less visible, but not entirely less traceable: users can hide the label, not erase the digital record.1