Google has launched a new online tool that lets people check whether an image, video, audio file, or text snippet was generated using its own artificial intelligence. The tool, called SynthID Detector, went live Tuesday during the company’s I/O 2025 event (though live now for just for select beta tester).
At its core, SynthID Detector works like a kind of reverse fingerprint scanner. Users can upload a piece of content, and the system looks for invisible markers left behind by Google’s AI models. If the tool picks up those digital watermarks, it flags the content — whether the whole thing or just part of it — as AI-made.
The timing of the launch isn’t random. AI-generated media along with a flood of misinformation have been flooding social platforms and search results at a pace that’s hard to track. Deepfake videos alone have exploded in number, rising more than fivefold between 2019 and last year, according to industry estimates. And it’s not just obscure corners of the web: some of the most popular Facebook posts in the U.S. last fall were clearly AI-built, based on public reports.
Still, this isn’t a silver bullet. SynthID Detector can only recognize content made with Google’s watermarking tech — which means anything generated by tools from OpenAI, Meta, or Microsoft flies under its radar. And even when Google’s own tools are used, the watermarking isn’t flawless. The company has said as much: in particular, it admits that AI-written text can sometimes slip past undetected if the content is altered or reformatted.
Even so, Google points to the scale of adoption so far. Since it introduced SynthID back in 2023, more than 10 billion pieces of media have been stamped with its watermark, according to internal numbers.
Whether that will be enough to keep pace with the growing flood of synthetic content is another question — but for now, Google is betting that more transparency is at least a step in the right direction.
Read next:
• Best AI Tools to Instantly Enhance Photos Without Editing Skills
At its core, SynthID Detector works like a kind of reverse fingerprint scanner. Users can upload a piece of content, and the system looks for invisible markers left behind by Google’s AI models. If the tool picks up those digital watermarks, it flags the content — whether the whole thing or just part of it — as AI-made.
The timing of the launch isn’t random. AI-generated media along with a flood of misinformation have been flooding social platforms and search results at a pace that’s hard to track. Deepfake videos alone have exploded in number, rising more than fivefold between 2019 and last year, according to industry estimates. And it’s not just obscure corners of the web: some of the most popular Facebook posts in the U.S. last fall were clearly AI-built, based on public reports.
Still, this isn’t a silver bullet. SynthID Detector can only recognize content made with Google’s watermarking tech — which means anything generated by tools from OpenAI, Meta, or Microsoft flies under its radar. And even when Google’s own tools are used, the watermarking isn’t flawless. The company has said as much: in particular, it admits that AI-written text can sometimes slip past undetected if the content is altered or reformatted.
Even so, Google points to the scale of adoption so far. Since it introduced SynthID back in 2023, more than 10 billion pieces of media have been stamped with its watermark, according to internal numbers.
Whether that will be enough to keep pace with the growing flood of synthetic content is another question — but for now, Google is betting that more transparency is at least a step in the right direction.
Read next:
• Best AI Tools to Instantly Enhance Photos Without Editing Skills
