Google SynthID: Global AI Content Detector Limits & Use

Google has released a dedicated public website for its SynthID content detector, centralizing the verification of synthetic media. According to Ars Technica, the platform allows users to check files for invisible watermarks embedded by multiple companies.

How does the new SynthID website work in practice?

Users can upload images, videos, and audio clips directly to the platform to check for AI generation. Ars Technica reports that the tool supports distinct watermarks from partners including OpenAI, Nvidia, Kakao, and soon Apple, alongside Google's own models like Gemini.

Despite opening access globally, Google imposes strict limits on the platform. Each user receives a daily quota of approximately 10 checks after logging in with a Google, OpenAI, or Apple account. According to engineers interviewed by Ars Technica, this restriction aims to stop malicious actors from developing tools to strip or bypass the watermarks.

What are the actual limitations for content verification?

The detector provides a simple yes-or-no response rather than detailed pixel-by-pixel breakdowns. Furthermore, TechCrunch notes that open-weight models, content without metadata, and watermarks from competitors using entirely different standards—like Meta—will not be flagged by the tool.