Google has launched a dedicated website, SynthID.com, allowing users to verify if images, videos, or audio files contain AI-generated watermarks. Previously, this functionality was restricted to internal testing or required queries through the Gemini chatbot, which often provided ambiguous results.
The updated detection tool now supports watermarks from various industry partners, including OpenAI, Nvidia, and Kakao, with Apple integration expected soon. This centralizes the verification process, eliminating the need to use separate detection tools for different AI platforms.
Users must sign in using Google, OpenAI, or Apple accounts to access the service. Google has implemented a daily usage quota of approximately 10 checks per user. Engineers state that these limits are necessary to prevent malicious actors from developing methods to remove or bypass SynthID watermarks.
The tool provides a simple confirmation of whether SynthID was used to create or edit media, rather than detailed pixel-level analysis. It does not identify content generated by models that use alternative watermarking technologies, such as Meta, or models that do not apply any watermarks at all.
Google notes that Gemini has already integrated SynthID into over 180 billion images and videos, as well as 240,000 years of audio content. The company acknowledges that the system is not exhaustive, as many AI-generated files remain outside the scope of current watermarking standards.
