Google announced on October 7 that the SynthID Detector is now open to everyone around the world, in English to start. Anyone can visit synthid.com, upload an image, video or audio file, and check whether it carries an AI-generation watermark. The tool was previously an early version shown at Google I/O last year, limited to journalists and researchers — this is its first real public debut.
The announcement came from Pushmeet Kohli, Google DeepMind's VP of Science and Strategic Initiatives, on the company's official Keyword blog. Google framed it as the next step in an "industry-wide effort to make AI content more transparent."
The biggest change in the new detector is coverage: beyond content from Google's own Gemini models, it recognizes watermarks from partners including OpenAI, NVIDIA and South Korea's Kakao. Apple has also confirmed it will join, with SynthID watermarks coming "soon" to images generated or edited with Apple Intelligence.
The supported formats are broad: JPG, PNG, HEIC and GIF for images; MP4, MOV and WEBM for video; MP3, WAV and FLAC for audio. Uploads happen right in the browser — no software to install.
SynthID doesn't work by "looking AI-generated." Instead, it weaves an imperceptible watermark into content at the moment of creation, and detection simply reads that birthmark. Since its 2023 launch, Google says it has watermarked more than 180 billion images and videos, plus the equivalent of 240,000 years of audio.
Google also revealed that built-in verification features in Search, the Gemini app and Chrome now regularly handle over 1 million requests a day. In other words, the system has been running in the background for a while — ordinary users are just now getting the front-door key.
Google goes out of its way in the detector's FAQ to stress: "This is not a general AI detector." It can only identify media from companies that have adopted SynthID technology. No watermark found does not mean a photo was taken by a human — it may simply have been made with tools outside SynthID's coverage.
Microsoft and Meta each run their own watermark standards, and the systems don't recognize each other. The world of AI-content verification remains fragmented: Google's ruler can't measure everyone's cloth.
For everyday readers, three use cases stand out: verify that viral "breaking news" image in the group chat before sharing it; confirm your own media assets are clean before publishing; and help older relatives spot AI scam calls — SynthID covers audio too.
Two things to watch next: when Apple's watermarking actually lands (which determines whether AI photos from iPhones can be verified at all), and whether Microsoft and Meta move toward SynthID. AI-generated content has become so abundant that telling real from fake by eye is nearly impossible. One ruler everyone recognizes beats ten "AI-smell" sniffers.