Google Opens SynthID Detector to Everyone Worldwide

Google has removed the wall between its SynthID Detector and the general public. The company said on October 7, 2026 that the checking tool, which had been limited to media professionals since an early version appeared last year, is now open to anyone online, starting with English. Pushmeet Kohli, who leads science and strategic initiatives at Google DeepMind, framed the move as part of giving people clearer context about what they see online so they can decide what to trust.

SynthID itself is older than the detector built around it. Google introduced the watermarking technology in 2023, embedding imperceptible marks into images, video and audio generated by its own models. Those marks survive common edits such as cropping, compression and filters, which is what lets the detector read them back out later. The new site lets a person upload a file and see whether any of those marks are present, rather than having to guess from how an image looks.

The detector does not stop at Google’s own models. The company said it can recognize content made with tools from partners including OpenAI, NVIDIA and Kakao, with Apple named as a coming addition. That breadth is the part of the announcement that separates this release from a simple product update: a single page now checks against watermarks set by several different AI labs rather than just the one that built it, assuming each partner continues to embed SynthID consistently and keeps that arrangement current as its own models change.

The numbers

• Google has applied SynthID watermarks to 180 billion images and videos since the technology launched in 2023.
• The company has also watermarked 240,000 years of audio content, a figure that signals how much synthetic audio the system has touched rather than how much any single listener will ever check.
• Built in verification features across Search, the Gemini app and Chrome already handle over 1 million requests daily, a volume the standalone detector page now adds to directly.

Google positions the expansion as routine safety infrastructure rather than a new product line, folding it in alongside its SynthID Bio work in synthetic biology and its existing verification features in Search and Chrome. The open question for a reader is less about Google’s own content and more about coverage: a watermark only helps if the tool that made the file actually applied one, and the detector can say nothing useful about content made with a generator that never adopted SynthID in the first place. That gap, not the technology itself, is where most AI-generated media will likely keep slipping through unflagged.