Artificial intelligence · Field note 025
Google opens its AI watermark detector to everyone
SynthID Detector can now inspect images, video, and audio for invisible marks from Google and participating AI companies.

The public now has another way to ask where a piece of media came from—provided its creator left the right kind of trace.
Google has opened SynthID Detector to everyone globally in English. The tool examines uploaded images, video, and audio for imperceptible watermarks embedded by Google’s generative systems and participating partners.
Those partners include OpenAI, NVIDIA, and Kakao, with Apple support described as coming soon. Google says its broader verification features in Search, Gemini, and Chrome already handle more than one million requests per day.
Detection by design
SynthID does not guess whether media looks artificial. It looks for a signal intentionally placed inside the file when the content is generated or edited. That makes the result more grounded than a detector trying to infer origin from visual artifacts or writing patterns.
The watermark is designed to remain invisible during ordinary use while surviving common transformations. Google says SynthID has been applied to more than 180 billion images and videos, along with the equivalent of 240,000 years of audio.
The detector began as an early-access tool for media professionals. Opening it to the public gives journalists, researchers, and everyday users the same basic verification route without requiring specialist access.
A useful answer with a firm limit
A negative result does not prove that a file was created by a person. SynthID Detector can only identify media carrying a supported watermark, so content from generators outside the participating network may pass through without a match.
That limitation makes the tool one part of a larger authenticity system, not a universal AI test. Its real promise depends on more companies adopting compatible marks—and on platforms preserving those signals as media moves around the web.