Google opens public SynthID detector to check content for AI

Google has opened public access to the SynthID Detector tool, which allows checking images, videos, and audio files for traces of artificial intelligence. The service works worldwide, but currently supports only the English interface language.

The technology is based on the SynthID system from Google DeepMind, which embeds an invisible watermark into digital content. Humans do not notice these changes, but detector algorithms can detect them.

The detector recognizes labels not only from Google but also from its partners: OpenAI, Nvidia, and Kakao. Apple will add support in the near future.

The scale of technology implementation is already huge:

• watermarks have been applied to more than 180 billion images and videos
• labeled audio amounts to about 240 thousand hours of playback
• similar checks are built into Search, Gemini, and Chrome, processing over 1 million queries daily

An early version of the service was launched in May 2025 with limited access for journalists and media specialists. Now the tool is available to all users.

There is an important nuance: the service searches exclusively for the SynthID label. If a content generator does not install this protection or uses unmarked open models, the detector will not be able to identify such a fake.