Google permits removal of visible watermarks from AI-generated content

Google has announced a significant adjustment to its approach to labelling artificial intelligence-generated media, permitting users to remove visible watermarks from images, videos and audio created by its Nano Banana, Omni and Lyria models. The change, which affects content produced through the Gemini app and Google’s Flow video editor, marks a pivot towards greater creative flexibility while maintaining underlying transparency mechanisms.

Josh Woodward, Google’s vice-president for Gemini, confirmed the update on X, stating that the company is striking a balance between creative control and safety. While the visible watermarks are now optional, Woodward emphasised that invisible SynthID watermarks and C2PA standard-related metadata remain active. This ensures that tools such as Gemini and Search can still identify whether an image or file was AI-generated, preserving a layer of content verification for users and platforms.

The new setting will be accessible via Settings > Media Watermark within the Gemini application and the Flow video editor, with support for Search expected to follow shortly. The feature is scheduled to roll out in the coming days. This development follows a broader industry trend where visible markers have faced criticism for reducing the utility of content for professional and creative workflows, even as the need to distinguish synthetic media from authentic material persists.

The move comes amidst heightened regulatory scrutiny and competitive dynamics in the generative AI sector. It follows Anthropic’s recent decision to include watermarks in text and files generated by its Claude model, a step taken to comply with European Union regulations. Unlike Anthropic’s regulatory-driven approach, Google’s adjustment appears framed around user experience and creative freedom, though it maintains robust technical safeguards against the complete erasure of content provenance.

In a parallel development aimed at supporting the developer ecosystem, Google is open-sourcing a new library called Credentio. This tool is designed to allow developers to embed local validation mechanisms directly into their applications, potentially standardising how third-party platforms verify the origin of AI-generated assets. The release of Credentio underscores Google’s strategy to provide infrastructure that supports transparency without imposing restrictive visible markers on end-user content.

These policy shifts occur against the backdrop of rapid adoption for Google’s AI tools. The Gemini app recently surpassed one billion monthly active users, placing it on par with OpenAI’s ChatGPT, which reached the same milestone in June. With 63 per cent of Gemini users engaging with the assistant via voice features, the volume of AI-generated content is escalating, making the distinction between synthetic and human-created media increasingly critical for digital ecosystems.