Google opens one site that checks media for AI watermarks
The tool finds media only from SynthID partners, and a no does not show that a person made it.
Google DeepMind released SynthID Detector for global use, and Google says it will add more partners. A user sends an image, audio or video to the SynthID site and gets a yes or no for media from partners that use SynthID. Before this, a user could use only Gemini to check media, and it found only Google content.
How it works
SynthID adds a mark that a person cannot see. The tool looks for this mark. These partners use it:
The evidence
Google gave no test results or error rates. SynthID is in the Best Inventions of 2026. 2 Reddit users are not sure about the tool. One tells that an easy step to remove noise from an image can remove the mark.
What is not known
Meta and Microsoft use a different mark, and the tool does not find their media. Large changes to the media can decrease the mark. The tool then does not find it. C2PA metadata is a record of the source of media. The record does not stay in screenshots.
Insight: Google builds a shared check, but two large rivals are outside
The site is for the partners that use SynthID, but not for Meta or Microsoft, which use a different watermark. This reads as a move to make its mark the standard, and each new partner adds value. For leaders: use more than one check, because a no is not a sign that a person made it.
Sources
Posted
