Google's SynthID Detector checks media for AI watermarks, with big blind spots
Google opened its web-based SynthID Detector to anyone who signs in. It scans images, video, and audio for SynthID, an invisible watermark that models from Google, OpenAI, Nvidia, and Kakao add to their output.
Why it matters
This gives you a quick first check on client-supplied media. The blind spots are big, though. It misses AI output that carries no SynthID watermark, including content from Claude, Grok, and Chinese open-weights models. It also cannot separate fully generated files from human work edited with AI.
Your moveRun one suspect image or clip through the detector. Treat a 'no watermark found' result as inconclusive, since it does not prove a person made the file.
Read the source