Google has opened SynthID Detector to the public at synthid.com, expanding access beyond the media professionals and researchers who had been able to use an earlier version. The free site is available worldwide in English and accepts image, video and audio uploads—but it checks for supported SynthID watermarks, not every kind of AI-generated media.
3
8
24
What the public detector does
SynthID is Google DeepMind’s watermarking technology. It embeds an imperceptible signal in AI-generated content so a compatible detector can look for it. Unlike information stored only in a file’s metadata, an embedded watermark is part of the content itself.
24
30
The public portal checks images, videos and audio for supported watermarks from Google, OpenAI, NVIDIA and Kakao. Apple support is planned, not yet listed as current coverage. Some reporting says ElevenLabs has committed to using SynthID, but that commitment alone does not establish that its media can currently be checked by the portal.
1
2
24
The launch broadens access to a system that Google says has been used to watermark more than 180 billion images and videos and 240,000 years of audio. Google also reports more than one million daily verification requests through Search, Gemini and Chrome. These are company-reported adoption and usage figures; they do not mean the public site can identify all AI-made media.
30
31
Access, language and sign-in
Anyone can visit synthid.com, but the site is currently available in English. Users must sign in with a Google, OpenAI or Apple account; the reported sign-in options do not include an alternative login method.
2
7
9
The public release is a meaningful change from the earlier limited-access portal. It also brings supported OpenAI watermarks into the check, extending verification beyond Google-made media.
3
8
How to interpret a result
A detected watermark is evidence that the file contains a supported SynthID signal. It is not a complete account of how the file was made: the result does not, by itself, establish whether the whole file was generated by AI or whether AI was used only to edit part of it. The public result also does not highlight which regions of an image carry the watermark.
8
A result with no detected watermark is not proof that a file is authentic or human-made. The detector only checks media from providers that have adopted SynthID, and edits can weaken or remove watermarks. Other AI systems may use different approaches—or no watermark the portal can detect.
19
24
Google’s detector site explicitly says it is not a general AI detector. The broader limitation is fragmented coverage: a tool can only verify signals it knows how to recognize. Treat the result as one useful check, not a final verdict about a file’s origin.
14
24