Google has launched a new public website, SynthID, enabling users to verify if media files, including images, videos, and audio, were generated by artificial intelligence. The tool now identifies watermarks from Google and its partner AI developers, expanding its reach beyond previous limited access.
Google has made its SynthID AI detection tool publicly available through a new website, allowing anyone to check images, videos, and audio files for invisible AI-generated watermarks. The expansion provides broad access to a technology previously limited to journalists, media professionals, and researchers. SynthID embeds imperceptible digital signatures into AI-generated media at the point of creation, designed to remain detectable even after common editing processes such as cropping or color filtering.
The SynthID Detector website now supports watermarks from Google's own AI models, including Gemini, as well as those from partners such as OpenAI, Nvidia, and Kakao. Google has indicated that support for Apple's AI models is expected soon, which would further broaden the tool's compatibility. Users are required to log in with a Google, OpenAI, or Apple account to use the service, and a daily quota of approximately ten checks is imposed to prevent misuse. Google has already applied SynthID watermarks to billions of images and videos and hundreds of thousands of years of audio content generated through its Gemini models.
Pushmeet Kohli, vice president of science and strategic initiatives at Google DeepMind, stated that the goal of SynthID is to provide users with "more context about the media you see online." The technology works by embedding these watermarks directly into the media's data, making them undetectable to human senses but readable by detection algorithms. Google developed SynthID starting in 2023, initially focusing on images before extending its capabilities to text, video, and audio. The text version of SynthID was made open source in 2024.
The public release of SynthID follows increased concerns about the proliferation of AI-generated content online, which can be difficult to distinguish from authentic media. Google has also integrated SynthID verification capabilities into other products, including Google Lens, AI Mode in Search, and the Chrome browser, with its integrated features processing over one million verification requests daily. While SynthID is designed to be robust, Google acknowledges that no detection signal is entirely foolproof and advises users to consider multiple sources of information when evaluating media. The tool identifies the presence of a SynthID watermark, indicating that media was made or edited with AI, but it cannot detect content generated by AI systems that do not use this specific watermarking standard.