Overview
- Google launched a public SynthID detector on Wednesday, October 7, 2026, at synthid.com so anyone can upload images, video, or audio for a watermark check.
- SynthID is an invisible watermark embedded in pixels or audio waveforms that signals content was produced or edited by participating AI models.
- The detector accepts common image, video, and audio formats and can recognize watermarks from Google plus partners including OpenAI, NVIDIA, and Kakao with Apple support expected soon.
- Users must sign in with a supported account and face a roughly 10-check daily quota, a limit Google says helps stop attackers from using results to build watermark-removal tools.
- A positive SynthID match is strong evidence a participating model made the media, but a negative result is inconclusive because many models don’t use SynthID and watermarks can be weakened or removed, so layered provenance systems like C2PA remain important.