SynthID Detector Goes Public as Watermarks Hit 180 Billion
Google has opened its SynthID Detector to everyone. Anyone can now upload an image, video or audio file and check whether it came from Google's AI models or from one of the partners using its watermarking system. Google also says the number of watermarked images and videos has passed 180 billion.
The tool has a clear limit. It does not tell you whether a file is AI-generated in general. It only tells you whether the file carries a SynthID watermark.
What the detector actually does
SynthID works by embedding a watermark into AI-generated content. People cannot see or hear it, but software can read it. The public detector scans uploaded files for that signal.
It supports four common formats:
- JPG and PNG for images
- MP4 for video
- MP3 for audio
The watermarks come from Google models such as Gemini, Veo and Lyria. Gemini is Google's main AI model family, Veo generates video and Lyria generates music. If a file was made with one of these, or with a partner tool that uses SynthID, the detector should flag it.
The catch: it only finds its own marks
This is the part most people will miss. SynthID is not a universal AI detector. If an image came from a generator that does not use SynthID, the tool has nothing to find. A clean result does not mean a file is real. It only means there is no SynthID watermark in it.
Google is not alone here. Other detection systems have the same limitation, and they usually only verify watermarks from their own provider. Each company checks its own output, and nobody checks everything.
Who else is on board
SynthID's reach goes beyond Google's own models. According to Google, OpenAI, Nvidia and Kakao already use SynthID for media. Kakao is a major South Korean internet company. Google also expects Apple to join soon.
OpenAI's case is split. It uses SynthID for images and video, but for text it built its own watermarking system. So even partners do not run every content type through Google's tools.
Already built into Google products
The public detector is not the only way to check content. SynthID detection already runs inside Google Search, the Gemini app and Chrome. Across these products, Google says it receives one million verification requests per day.
Then there is the headline number. Google says more than 180 billion images and videos carry SynthID watermarks so far. That shows how much synthetic media Google and its partners are already putting out. It also shows how much content could, in theory, be traced back to its source.
Why It Matters
For readers, the practical lesson is simple: treat the SynthID Detector as one check among several, not as a verdict. A positive result is useful evidence that a file came from Google or a partner. A negative result tells you very little about where it came from.
This launch also fits a broader pattern. Synthetic media keeps getting more convincing. We covered one AI avatar that fooled 48% of people on short calls. Meanwhile, Google itself keeps making image generation cheaper. More output at lower cost makes provenance tools more important. Fragmentation, however, makes them less reliable.
The partner list suggests SynthID could become a shared standard for media, at least among large providers. It is worth watching whether Apple actually joins. Another question is whether rival systems start reading each other's watermarks. Until that happens, most detection results are likely to come with an important caveat.