# Google now lets users toggle off visible watermarks from AI-generated content > Source: > Published: 2026-08-14 16:45:59+00:00 Photo: Vgairlinewiki1 / Wikimedia Commons / CC BY 4.0 (https://creativecommons.org/licenses/by/4.0) # Google now lets users toggle off visible watermarks from AI-generated content The sparkle logo is going optional, but invisible tracking stays baked into every image, video, and audio clip that Gemini produces. Google just made a subtle but meaningful concession to the millions of people using its AI tools to generate images, videos, and music. Users can now toggle off the visible “sparkle” watermark that appears in the bottom-right corner of content created with Gemini and Flow, Google’s AI video generator. The catch: invisible watermarks aren’t going anywhere. Every piece of AI-generated content will still carry SynthID invisible watermarks and C2PA metadata embedded in the background, whether you want them there or not. ## What’s actually changing The new setting lives in the Gemini app under Settings > Media Watermark. The toggle also works in Flow, Google’s dedicated AI video tool. Josh Woodward, VP of Google Labs, Gemini, and AI Studio, framed the update as a balance between user preferences and content integrity. The visible branding is now optional. The invisible verification layer is not. That invisible layer does real work. SynthID embeds imperceptible signals into generated media that survive screenshots, crops, and format conversions. C2PA Content Credentials, meanwhile, attach metadata about how and when the content was created. Together, they let anyone with the right tools, like the SynthID Detector, confirm whether a piece of media came from an AI system. Not everyone gets the toggle, though. Work and school accounts may not see the option at all. Users in India, South Korea, and Vietnam need an AI Ultra subscription to access it. ## The scale of SynthID so far SynthID has embedded invisible watermarks in over 100 billion images and videos since its inception. On the audio side, the system has processed the equivalent of 60,000 years of sound. By 2026, Google intends to integrate the technology into Google Search and Chrome. That expansion will also involve partnerships with other major players, including OpenAI. ## Why the visible watermark mattered, and why it’s going away Visible watermarks serve a simple purpose: they tell anyone looking at a piece of content that a machine made it. No special tools needed. No metadata inspection. Just a small logo in the corner that says “this didn’t come from a human.” The problem is that visible watermarks also make AI-generated content look unfinished. For creators using these tools in professional workflows, a branded sparkle logo in the corner is about as welcome as a stock photo watermark on a client presentation. The invisible watermark approach tries to have it both ways. Casual viewers won’t know the content is AI-generated unless someone tells them. But platforms, fact-checkers, and forensic analysts can still verify provenance using SynthID detection or C2PA credential checks. ## What this means for the digital content landscape The EU AI Act includes provisions around synthetic content labeling. Several US states have passed or proposed laws targeting deepfakes, particularly in election and adult content contexts. For content platforms, the shift puts more pressure on detection infrastructure. If visible watermarks become rare, platforms like YouTube, Instagram, and X will need to rely more heavily on invisible signals like SynthID and C2PA to identify and label AI content in their feeds. Google’s plan to bake SynthID into Chrome and Search by 2026 could make that detection nearly automatic, at least within Google’s own ecosystem. OpenAI, Meta, and Adobe have all adopted C2PA standards to varying degrees. Google’s expansion of SynthID alongside C2PA creates a dual-layer system that competitors may feel pressure to match. **Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).