Overview
- The detector, opened Wednesday, October 7, 2026, is available worldwide on a dedicated website and can recognize SynthID marks placed by participating partners such as OpenAI, NVIDIA, Kakao, and others.
- SynthID embeds an invisible signal inside content so images and video carry pixel-level markers and audio carries a waveform signal that a detector can read to confirm a participating model produced the file.
- Google requires users to sign in with a Google, OpenAI, or Apple account and enforces a daily quota of about 10 checks per user to stop people from running bulk queries that could be used to build watermark-removal tools.
- The detector can only confirm files that carry a SynthID watermark from participating models, so a negative result does not prove a file is human-made and watermarks can be absent or removed by attackers.
- Google says its Gemini service has already applied SynthID at huge scale—over 100 billion images and videos and 60,000 years of audio—and planned extensions such as SynthID Bio and Apple iOS integration could broaden use but leave adoption and removal-resistance as open questions.