Google is changing how it handles visible AI watermarks on media produced with its Gemini tools. Up until now, the company had made a point of ensuring that images, video and other media created with AI carried visible markings to help users identify them. With a recent update, Google is giving users more control: visible watermarks added by Gemini can now be turned off.
This change affects various types of AI-generated media, including photos, videos and music created through Gemini. Even when the visible watermark is disabled, Google says that invisible markers will remain. Specifically, content will continue to include SynthID markers and C2PA metadata embedded in the files. Those invisible markers are intended to preserve a degree of provenance and enable verification via tools that can read that metadata, but they are not obvious to casual viewers who do not look for them.
To manage the visible watermark, Google has provided an option inside the Gemini app. Users who prefer not to see the watermark can go to Settings > Media watermark within Gemini and disable the visible mark. Google has said this toggle will roll out to both the Gemini app and Google Flow in the coming days, giving users consistent control across those experiences.
Even with the visible watermark option, Google emphasizes that transparency features remain. Invisible SynthID watermarks and C2PA metadata will be preserved so that, when necessary, content can still be inspected to determine whether it was AI-generated. That means people who need to verify authenticity can rely on Gemini or Search to check whether a specific image or clip originated from an AI model.
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.
There are a few important caveats. In countries or regions where local law requires visible labeling of AI-generated content, Google will continue to display the Gemini-visible watermark regardless of user settings. That exception ensures the company complies with legal obligations where transparency mandates are in place.
Critics of removing visible watermarks argue that optional marking reduces the effectiveness of consumer-facing signals designed to curb misinformation and deception. Visible watermarks act as an immediate cue to viewers that media may not be authentic, and making them optional could make it easier for manipulated or synthetic content to spread without obvious warning. Proponents of the change counter that visible watermarks can be intrusive for creators who want a clean result for legitimate creative work and that embedded metadata still preserves the ability to verify provenance when necessary.
Google frames the update as a compromise between creative flexibility and the need for accountability. By keeping invisible provenance markers in place while enabling users to toggle visible watermarks, the company is attempting to support creators who prefer an unmarked aesthetic while retaining technical mechanisms for later verification. The company also points to built-in tools in Gemini and Search that can be used to interrogate media and confirm whether it was produced by AI, offering a route to transparency without forcing a visible overlay on every piece of content.
For users and organizations that rely on visible cues to assess content authenticity, the change means they will need to adapt workflows. That could include using platform tools to check metadata, relying on third-party verification services, or applying internal policies that require visible labeling despite the user setting. For casual readers and viewers, invisible markers are unlikely to alter behavior immediately, since most people do not routinely inspect file metadata or use verification tools unless a piece of content raises doubts.
In short, Google’s update gives individuals more control over whether AI media displays a visible Gemini watermark, while retaining embedded methods of provenance and adding exceptions where law requires visible notice. The balance it seeks aims to serve both creators who want an uncluttered creative output and institutions that need reliable ways to confirm whether content was created with AI.