Google is changing how it labels AI-generated media by making visible watermarks optional for some of its creative AI outputs. The update applies to AI generations including images, videos, and songs, while keeping other identification systems in place.
The company announced on Friday that users will be able to turn the visible mark on or off. The move gives creators more control over how finished media appears, without removing the invisible signals Google says are used to help identify AI-generated content.
What Google Is Changing
The new setting lets users remove a visible watermark from AI generations. According to Josh Woodward, Google’s VP for Gemini, the toggle will be available for Nano Banana, Omni, and Lyria models.
The setting will appear in Gemini and Google’s video editor, Flow. Support for Search is coming soon.
Once the feature is available, users will be able to go to Settings > Media Watermark to turn the visible mark on or off. The feature is rolling out in the coming days.
For people making professional or creative work, the change matters because a visible watermark can affect how media is presented, edited, or used in a finished project. Google’s update recognizes that visible labeling can be useful for transparency, but can also limit the practical value of an AI-generated image, video, or song.
What Is Staying In Place
Google specified that removing the visible watermark will not affect the invisible SynthID watermark or C2PA standard-related metadata. In other words, the public-facing mark may become optional, but Google is not removing every form of AI media identification.
Woodward described the decision as an attempt to balance creative control with safety. He wrote: “We’re striking a balance here between creative control and safety: while the visible watermarks are now optional, invisible SynthID watermarks and C2PA metadata are still being used for transparency. So you can still use Gemini or Search to see if an image was AI-generated,”
That distinction is central to the update. A visible watermark is meant to be seen directly on the media. Invisible SynthID watermarks and C2PA metadata are different signals that can support transparency without placing a visible mark over the output itself.
The result is a two-track approach. Users get more control over the appearance of generated media, while Google keeps mechanisms connected to identifying AI-generated content.
Why The Update Matters
The change reflects a broader tension around AI-generated media. Labels can help people understand when content was made by an AI system. At the same time, a visible watermark can make an otherwise usable asset less suitable for professional or creative work.
Google’s decision suggests that AI media labeling is moving beyond a single visible mark. Instead, the company is leaning on a combination of user-facing controls and less visible transparency systems.
For creators using Gemini or Flow, the most immediate effect is practical. They will be able to decide whether the visible watermark belongs on a particular output. For people checking whether something was AI-generated, Google says Gemini or Search can still be used to see if an image was AI-generated.
The update applies across multiple kinds of generated media. The source specifically names images, videos, and songs, and also names Nano Banana, Omni, and Lyria as models tied to the toggle.
Developer Tools And The Wider Context
Google is also open-sourcing a new library called Credentio. The purpose is to let developers embed a local validation mechanism in their apps.
That developer-facing step fits with the same theme as the watermark toggle: AI content identification is not only a platform setting, but also something other apps may need to validate locally.
The announcement also arrives after Anthropic’s hotly debated move to include a watermark in text and files generated by Claude to comply with EU regulations. The source does not describe Google’s change as a direct response, but it places the update in the same wider debate over how AI-generated material should be marked.
For now, Google’s position is clear from the announced rollout. Visible watermarks are becoming optional in supported products, but invisible SynthID watermarks and C2PA metadata remain part of the company’s transparency approach.