Google's SynthID Detector Opens to Everyone Worldwide

Google announced on October 7, 2026, that its SynthID Detector is now open to everyone around the world, in the widest release yet of a tool that can spot invisible watermarks in AI-generated media. The announcement, published on the company's official blog The Keyword and written by Google DeepMind executive Pushmeet Kohli, says people can now visit synthid.com and check whether an image, video, or audio clip carries the hidden signature of an AI model built by Google or one of its partners, including OpenAI, NVIDIA, and Kakao. Apple is expected to join the program soon, according to the post.

The detector works simply: a person uploads a file, and the portal reports whether it finds a recognized SynthID watermark embedded at the moment of creation. The site accepts everyday formats such as JPEG and HEIC images, MP4 and WEBM video, and audio files ranging from WAV to M4A. Google is explicit about the tool's boundaries, noting in the site's FAQ that it is not a general AI detector. It can only identify media watermarked by companies that have adopted the technology, so a clean result does not prove a file is authentic. It only means no recognized watermark was found.

Three years of invisible watermarking

SynthID has been quietly marking AI output since its launch in 2023, when Google DeepMind first described the system for embedding imperceptible signals into generated images without reducing their quality. Since then, the company reports that it has watermarked more than 180 billion images and videos, along with audio content totaling roughly 240,000 years of listening time. That scale makes the watermark layer one of the largest provenance systems ever applied to synthetic media, and the detector is the first time the public has been able to query it directly.

An early version of the detector went to a limited group of media professionals in 2025, giving journalists and researchers a way to verify AI-generated content for their work. This week's expansion brings the same check to the general public, and it arrives alongside verification features already built into Search, the Gemini app, and Chrome. Google says those built-in features now handle more than a million verification requests every day, a sign that checking the origin of media is becoming a routine habit rather than a specialist task.

What the tool cannot prove

The limits matter as much as the capability. SynthID only detects content watermarked by participating models, and it says nothing definitive about files created outside that circle. A video generated by a model that does not use the technology will pass through the detector silently, which means a negative result is not a certificate of authenticity. Kohli acknowledged the gap in the announcement, writing that as AI tools become better and easier to use, having clear context about the content people encounter online helps them make informed decisions about what to trust.

That context is arriving at a moment of real pressure on information quality. Research cited by technology outlets has found that several of the most-viewed posts on major social platforms last year were created with AI tools, and fabricated images and audio have already surfaced in political campaigns, financial scams, and news events. Independent coverage of the launch by Unite.AI notes that the public release follows the early prototype that was restricted to media professionals last year. Watermarking does not solve that problem on its own, but a shared standard backed by Google, OpenAI, and other large model builders gives newsrooms, courts, advertisers, and ordinary readers a common reference point instead of a patchwork of incompatible checks.

For readers, the practical move is straightforward: run suspicious media through the detector before sharing it, look for the verification indicators already appearing in Search and Chrome, and treat a no-watermark result as inconclusive rather than as proof that something is genuine. Newsrooms can fold the check into their verification workflow, platforms can point users to it when a clip goes viral, and anyone who receives a forwarded video of uncertain origin can get an answer in seconds. The technology is a supplement to skepticism, not a replacement for it. With the portal now open worldwide, the question is no longer whether people can check where synthetic media came from, but whether they will bother to do it.