Google has launched SynthID.com, a public website that lets users verify whether an image, video, or audio clip was generated using AI. This follows a test period for journalists, media professionals, and researchers during Google I/O the previous year.
What SynthID Can Do
The platform accepts a wide variety of file types. For images, it supports formats like JPG, PNG, WEBP, AVIF, HEIC, TIFF, GIF, and others; videos in MP4, MOV, and WEBM; and audio in WAV, MP3, OGG, FLAC, AAC, and M4A. Users upload their media, and SynthID tells them whether it was generated by an AI model that uses Google’s watermarking technology.
Where the Watermarks Come From
SynthID is the mechanism Google introduced in 2023 to watermark AI-generated content. Media produced by Google’s generative models—Nano Banana, Veo, Lyria—and by tools like Gemini, Flow, ProducerAI, and Vids now include the watermark. Other companies including OpenAI, Nvidia, and Kakao also support SynthID watermarking, and Apple plans to adopt it soon.
The tool isn’t just standalone. SynthID verification is built into Google’s Gemini app and in Google Chrome. Google reports around one million daily verification requests through its systems.
Limitations and the Landscape
While Google, Microsoft, Meta, and others have developed watermarking and detection systems, none are perfect. Even tools designed by the same companies whose models created the content sometimes fail to correctly identify AI-generated media.
SynthID’s open availability marks a shift toward more transparency in identifying generative media. But it’s also a reminder that watermark-based detection has its limits—especially as generative AI becomes more sophisticated.
SynthID’s public launch simplifies verification of AI content. By making detection accessible to all, Google aims to combat misinformation and misuse stemming from deepfakes and synthetic media. This could set new norms for accountability in AI content generation.
What to watch: how reliable SynthID remains when faced with maliciously altered or adversarially prepared media—and whether other major players adopt similar open verification tools. If emerging products ignore watermarking, the risk of undetectable synthetic media could grow.
