Google has quietly turned one of its most interesting trust-and-safety tools into a public utility. On October 7, 2026, the company opened its SynthID Detector to everyone worldwide in English at synthid.com, letting any internet user check whether a file was generated by AI. Until now, the tool was restricted to journalists, media professionals and researchers, having first been previewed at Google I/O 2025.
The idea is simple. You upload a file — an image in formats like JPG, PNG, HEIC or GIF; a video in MP4, MOV or WEBM; or an audio clip in MP3, WAV or FLAC — and the detector scans it for SynthID, Google's invisible watermarking technology. These watermarks are embedded in content when AI models generate it, and unlike visible labels, they survive ordinary edits like cropping or compression. If the scan finds a matching watermark, it tells you the file was AI-made and often identifies which model produced it.
Notably, the detector is not limited to Google's own output. It recognizes watermarks from Google's models and tools as well as partner companies including OpenAI, NVIDIA and Kakao, and Google says support for Apple is coming soon. That cross-industry reach is significant: a watermarking system is only as useful as its adoption, and a checker that spans the biggest AI labs gives the whole effort real teeth.
The scale behind it is staggering. Google says that since launching SynthID in 2023, it has watermarked more than 180 billion images and videos, along with 240,000 years' worth of audio. The company also reports that users already make roughly 1 million verification requests per day through built-in checks in Search, the Gemini app and Chrome.
There are limits, and Google is upfront about them. You must sign in to use the portal, and each user is limited to around 10 checks per day — a cap Google says helps prevent bad actors from probing the system to build watermark-removal tools. More importantly, this is not a general-purpose AI detector. It can only find SynthID watermarks, so the absence of a watermark proves nothing: a file without one may simply never have been watermarked. The tool also cannot distinguish AI-created content from AI-edited content, since both carry watermarks.
Still, as deepfakes flood social feeds and confuse even careful viewers, a free official checker that anyone can use is a genuine step toward accountability. Just remember what it can and cannot tell you.
Comments
Post a Comment