Google Will Now Let You Remove the Visible Badge on Its AI-Generated Images and Videos
Google now lets users remove the small visible badge from AI-generated images, videos, and music created with its AI tools. A new "Media watermark" toggle in Gemini and Google's AI video generator, Flow, turns off the sparkle icon that normally appears in the bottom-right corner of content made with Google's Nano Banana and Omni models (The Verge, 2026-08-14).
But turning off the visible badge does not remove all traces. Google keeps two hidden marking systems embedded in the content. Josh Woodward, a vice president who oversees Google Labs, Gemini, and AI Studio, confirmed that both systems stay in place regardless of the toggle setting (The Verge, 2026-08-14).
The first is called SynthID. It is Google's own technology for hiding a digital signature inside an image, video, or audio file. You cannot see it, but the right software can detect it and confirm the content was made by AI. SynthID is already used in Gemini Omni video outputs, where each generated video carries an embedded signature that can be independently verified (Google Blog). The second is called C2PA, a standard from the Coalition for Content Provenance and Authenticity. C2PA works like a label attached to the file itself, recording where the media came from and how it was produced. The two systems complement each other.
Anyone can check whether a piece of content carries these hidden markers by asking Gemini or Google Search whether it is AI-generated (The Verge, 2026-08-14). Google's Gemini app added a dedicated AI image verification feature in November 2025, letting users check whether content was created or edited by AI (Google Blog, 2025-11-20). For organizations, Google launched an AI Content Detection API on Google Cloud's Gemini Enterprise Agent in May 2026 to help them identify AI-generated media at scale (Google Blog).
Google plans to bring the Media watermark toggle to Google Search, though it will not be available in countries that require visible watermarks by law (The Verge, 2026-08-14). Google's support documentation confirms the visible watermark covers all media types produced in Gemini apps, including images, videos, and music tracks, each of which also carries a hidden digital watermark for identification purposes (Google Support; Google Support).
Anthropic, another major AI company, announced it is also applying invisible watermarks to both AI-generated text and images (The Verge, 2026-08-14). The fact that two of the biggest AI companies are making the same move at the same time suggests the industry is settling on hidden markers as the main way to identify AI-generated content, rather than visible badges.
Users have been trying to get around visible watermarks for some time. A Gemini support thread from March 2026 documented Ultra subscribers removing the visible watermark by logging out of all Google accounts and clearing browser cache and site data, a workaround that existed before the official toggle (Google Support, 2026-03-05). Earlier, in March 2025, TechCrunch reported that users were using Google's Gemini model itself to strip watermarks from existing images, based on social media reports (TechCrunch, 2025-03-17).
The tension behind this change is straightforward. A visible badge tells anyone who sees the content that it was made by AI. But it also gets in the way for people using AI-generated content professionally, where a sparkle icon in the corner of a marketing image or a video clip is often unacceptable. Google's decision to let users turn off the visible mark, while keeping the hidden markers underneath, is an attempt to solve that problem. The trade-off is that checking whether something is AI-generated now requires a tool, rather than just looking at it.
Whether that approach works depends on how well the hidden watermarks hold up in the real world. SynthID is designed to survive common changes like compression, cropping, and color adjustment, but Google has not publicly shared how well it holds up against every possible way someone might try to remove it. C2PA metadata can be stripped from a file entirely if someone re-encodes the media without preserving the provenance information. Neither system offers a guarantee as strong as, say, a locked digital receipt that proves exactly who created a file and when.
In my view, the real question is not whether hidden watermarks can be defeated by a determined adversary. It is whether the tools for detecting them — the verification APIs, the in-app checks, the platform-level enforcement — become reliable and accessible enough that hidden markers can truly replace visible ones. Google has built most of that stack already: SynthID for embedding markers, the AI Content Detection API for large-scale identification, and in-app verification for everyday users. Anthropic's parallel move suggests the major AI companies are making the same bet. The concern is that detection tools may stay accessible only to platforms and their enterprise customers, while ordinary people lose the visible cue that required no special software at all.
On balance, Google's approach is a reasonable bet that behind-the-scenes provenance can do the job that visible badges could not. The trade-off is real, and the outcome will depend on how well it all works in practice rather than on the announcement itself.
