Google lets users turn off visible AI watermarks
Google announced on Friday that it will let users remove the visible watermark stamped on media made with its AI tools, covering images, videos and songs. The change does not touch the invisible SynthID watermark or the C2PA standard metadata attached to the same files; those keep running in the background regardless of the toggle.
Josh Woodward, Google's VP for Gemini, announced the change in a post on X. He said the toggle will cover three model families: Nano Banana, Omni and Lyria. The setting will be available first in Gemini and in Flow, Google's video editor, with support for Search described as coming soon. Once the rollout reaches a user, the control sits at Settings > Media Watermark, where the visible mark can be switched on or off. Google says the feature is rolling out over the coming days.
Woodward framed the move as a balance between creative control and safety: visible marks become optional, but the invisible SynthID watermark and C2PA metadata keep working underneath, so Gemini or Search can still be used to check whether an image was AI-generated.
Alongside the toggle, Google is open sourcing a new library called Credentio, meant to let developers embed a local validation mechanism directly in their own apps rather than relying on Google's tools to check provenance.
The announcement lands soon after Anthropic's own watermarking move drew criticism: Anthropic added a watermark to text and files generated by Claude to comply with EU regulations, a step the article describes as hotly debated.
Key facts
- Google will let users remove the visible watermark from AI-generated images, videos and songs, without affecting the invisible SynthID watermark or C2PA metadata.
- The toggle covers three model families: Nano Banana, Omni and Lyria, announced by Josh Woodward, Google's VP for Gemini, on X.
- The setting rolls out first in Gemini and in Flow, Google's video editor, with Search support coming soon, controlled at Settings > Media Watermark.
- Google is open sourcing a new library, Credentio, so developers can embed a local validation mechanism in their own apps.
- The move follows Anthropic's hotly debated decision to watermark text and files from Claude to comply with EU regulations.
Why it matters
Visible watermarks make AI output easy to spot at a glance but also make it less usable for professional and creative work, since the mark sits on top of the image, video or track itself. Google is now separating the two jobs: the visible mark becomes a user choice, while the invisible SynthID watermark and C2PA metadata keep doing the provenance work underneath. That lets the same file be both clean enough for creative use and still checkable through Gemini or Search. The timing also matters: it lands right after Anthropic's own watermarking move for Claude, taken to comply with EU rules, drew heavy criticism, giving the industry two different answers to the same labeling problem within days of each other.
Who it affects
Anyone generating media with Google's Nano Banana, Omni or Lyria models, since those are the three families Josh Woodward named as getting the toggle. In practice that means users of Gemini and of Flow, Google's video editor, first, with Search gaining the same control soon. It also affects developers building on top of AI-generated content, since Credentio, the newly open sourced library, is aimed at letting them add their own local check for whether a file carries the invisible watermark or C2PA metadata, rather than depending solely on Google's own tools.
How to use it
The feature is rolling out over the coming days rather than all at once. Once it reaches a given account, the control lives at Settings > Media Watermark, where the visible mark can be turned on or off for generations made with Nano Banana, Omni and Lyria. It is available first inside Gemini and inside Flow; Search support is described only as coming soon, with no date given. Developers who want their own provenance check outside Google's apps can use Credentio, the open source library Google is releasing alongside the toggle, to embed a local validation mechanism.
How solid is it
The claims trace to a single named source: Josh Woodward, Google's VP for Gemini, speaking in his own post on X, which TechCrunch reports directly rather than through a Google blog post or press release. That makes the announcement a first-party company statement rather than independently verified reporting, though it comes from a named executive on the record rather than an anonymous or unnamed spokesperson. No rollout percentage, region list or exact date is given beyond "the coming days" and "coming soon" for Search.
Risks and caveats
Removing the visible mark means a casual viewer loses the immediate visual cue that a piece of media was AI-generated; verifying it now requires actively checking through Gemini or Search rather than seeing it at a glance, which raises the bar for catching AI content in contexts where nobody thinks to check. The article gives no technical detail on how Credentio's local validation mechanism actually works, no calendar date for the Friday announcement beyond the day of the week, and no date for when Search support arrives. It also gives no detail on Anthropic's EU watermarking move beyond calling it hotly debated, so the comparison is asserted rather than spelled out.
“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.”
— Josh Woodward, Google's VP for Gemini