Google now lets users hide AI-generated content watermarks

Google is rolling out the ability to disable visible watermarks on media created by its AI systems. Users can now toggle off a "Media watermark" setting in Gemini and the Flow video generator, removing the distinctive sparkle icon that previously appeared in the corner of generated content.

What's changing#

The update allows users to hide the visible watermark that Gemini's Nano Banana and Omni models place on images, videos, and music. This addresses a growing trend: many users have already been manually removing these markers, and competing services like OpenAI and Meta's AI tools don't apply visible watermarks at all.

Invisible tracking remains#

Despite allowing visible watermark removal, Google is keeping invisible identification layers in place. All AI-generated content will continue to carry SynthID watermarks and C2PA metadata embedded within files. This means users can still verify whether something is AI-generated by asking Gemini or Google Search directly.

Industry pattern#

The shift aligns with how other major AI providers handle content marking. OpenAI and Meta both use invisible watermarking systems rather than visible ones, while Meta has developed its own Content Seal standard. This week, Anthropic announced it's applying invisible watermarks to its Claude AI's text and image outputs.

Broader rollout planned#

Google intends to extend the watermark toggle to Search as well, though it won't activate this feature in jurisdictions that legally mandate visible watermarks on AI content.