Google Launches SynthID Detector Worldwide So Anyone Can Check if Photos, Videos and Audio Were Made by AI

MOUNTAIN VIEW, Calif. — Google has opened its artificial intelligence detection tool to the public worldwide, letting anyone check whether an image, video or audio file was created with AI tools from Google and several major industry partners, including OpenAI.

The SynthID Detector website, launched globally in English on Tuesday, scans uploaded media for SynthID, an invisible digital watermark that Google and its partners embed in content generated by their AI models.

The portal had been limited to a small group of testers, including journalists and researchers, for more than a year.

"Today, we're expanding SynthID Detector in partnership with @OpenAI, @NVIDIA, Kakao & soon @Apple as part of an industry-wide effort to make AI-generated content more transparent," Google said in a post on X.

Detecting more than Google's AI

The biggest change is that the tool is no longer limited to Google-made content.

In addition to media created with Google's own AI models, the detector can now identify watermarked content generated by tools from OpenAI, chipmaker Nvidia and South Korean tech company Kakao. Google said support for images created with Apple Intelligence is coming soon.

Until now, verification was fragmented. Google's Gemini chatbot could only detect watermarks made by Google's AI tools, while OpenAI's verification website could only detect SynthID signatures from OpenAI products, according to Android Authority.

The new portal brings those checks together in one place.

How it works

Users can upload an image, video or audio file to the website, which then scans the file for the SynthID watermark. The tool supports dozens of common file formats, including JPG images, MP4 videos and WAV audio files.

Unlike traditional metadata labels, which can easily be stripped away when a file is saved, edited or uploaded to social media, SynthID is embedded directly into the content itself.

For images and video, the watermark is woven into the pixels in a way that is imperceptible to the human eye. For audio, it subtly alters the waveform in a way people cannot hear but software can detect.

When Google first introduced the portal in May 2025, it said the tool could highlight which specific parts of a piece of content were most likely to carry a watermark, such as particular segments of an audio track or areas of an image.

Massive scale

Google says SynthID has been embedded in more than 180 billion images and videos, along with the equivalent of 240,000 years of audio, according to Digital Trends.

The detector is already handling about 1 million requests a day, according to reports on the launch.

That growth reflects how quickly the technology has spread. When Google announced the detector portal last year, it said more than 10 billion pieces of content had been watermarked with SynthID.

Important limitations

Google and experts caution that the tool has significant limits.

The detector can only identify content that carries a SynthID watermark. That means AI-generated media from companies that do not use the technology, such as tools from Meta or Microsoft, will not be flagged as AI-generated.

As a result, a negative result does not prove that a file is authentic. An image without a SynthID watermark could still have been made with AI.

The tool also cannot distinguish between media that was entirely created by AI and real photos or videos that were only edited with AI tools.

Watermarking systems can also struggle with heavily edited, cropped or compressed media, and Google has previously acknowledged that SynthID is not foolproof against extreme manipulation or deliberate attempts to remove it.

Fighting deepfakes and misinformation

The global rollout comes amid growing concern about the spread of realistic AI-generated content, including deepfake videos, fake photos and cloned voices used in scams and political misinformation.

Modern AI voice tools can laugh, whisper, pause naturally and convey emotion convincingly enough to fool listeners, Digital Trends noted, making detection increasingly difficult for ordinary people.

When Google first unveiled the detector, Pushmeet Kohli, vice president of science and strategic initiatives at Google DeepMind, said advances in generative AI were making it possible for people to create content in entirely new ways.

"Content transparency remains a complex challenge," Kohli said at the time.

Regulatory pressure

The launch also comes as governments around the world move to require labeling of AI-generated content.

The European Union's AI Act includes provisions requiring that synthetic content be identifiable, and several U.S. states have considered or passed laws requiring disclosure of AI-generated media, particularly in political advertising. China has also introduced rules requiring AI-generated content to be labeled.

By building an industry coalition around SynthID, Google is positioning its watermarking technology as a potential global standard.

Competing standards

SynthID is not the only effort to track AI-generated content.

A separate industry group, the Coalition for Content Provenance and Authenticity, or C2PA, has developed "content credentials," a metadata standard that records where a piece of media came from and how it was edited. Google, OpenAI, Microsoft, Adobe and others are members of that group.

Experts say no single approach is likely to solve the problem on its own, and that watermarking and provenance tools are most effective when combined with media literacy and other verification methods.

How to use it

The SynthID Detector is free to use and available on the web. Users upload a file, and the portal returns a result indicating whether a SynthID watermark was detected.

Google has also made its SynthID text watermarking technology available as open source, allowing developers to add watermarks to AI-generated text.

The company has not said when the detector will be available in languages other than English.