Your mom sends you a Facebook photo of something incredible and asks whether it is real. Maybe it is an animal that looks a little too perfect, a celebrity doing something outrageous, or a disaster scene that has everyone in the comments freaking out. You suspect AI, but pointing out a weird hand is hardly a reliable verification method.
Google is giving her another way to investigate. The company has opened its SynthID Detector to everyone globally in English, allowing people to check images, videos, and audio files for invisible watermarks from supported AI tools.
According to Google’s announcement, the detector recognizes watermarks from Google and partners including OpenAI, NVIDIA, and Kakao. Apple will join them soon, although Google does not explain what its participation involves or provide a launch date. That leaves us without confirmation of any specific integration into iOS, Photos, or Apple Intelligence.
Google launched SynthID in 2023 and introduced an early version of the detector for media professionals last year. Opening access to the public brings that capability to people who want to investigate something a friend shared, without needing access to a professional verification service.
The company says it has watermarked more than 180 billion images and videos, along with 240,000 years of audio content. Those company-reported figures offer some perspective on the enormous volume of synthetic media already circulating.
SynthID embeds signals directly into AI-generated content that people cannot ordinarily see or hear. The detector searches an uploaded file for those signals, providing evidence of AI involvement when it finds a supported watermark. Google also offers verification through Search, the Gemini app, and Chrome, where it says these features regularly handle more than 1 million requests daily.
The limitation is straightforward: finding no watermark does not prove something is authentic. An unsupported generator could have created the file, and Google has acknowledged that adversarial actors can manipulate or circumvent watermarking. The public announcement also lists images, video, and audio, without promising text detection.
Even a successful detection leaves questions about the surrounding post unanswered. An AI image might accompany a clearly labeled joke, while a genuine photograph might carry a completely fabricated caption. Knowing how someone created the media helps, but checking the claim still requires context.
I can see myself sending this to a relative who regularly forwards questionable Facebook photos. It gives people something more concrete than arguing over whether an image looks fake, provided they understand the limits. Your mom may get useful evidence about that suspicious photo, but she should still check the story attached to it before hitting Share.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz
