Google's AI watermark detector goes public, now checks all major platforms

Google has made it simpler to detect artificially generated content by opening up its SynthID watermarking detection system to the general public. Starting today, visitors to a new dedicated website can upload media to check whether it bears invisible markers left by AI tools from Google, OpenAI, Nvidia, Kakao, and soon Apple.

How SynthID works#

SynthID operates by embedding a hidden signal into the fabric of digital media—encoded in image pixels, video frames, and audio waveforms. Google's Gemini models alone have integrated this watermark across more than 180 billion images and videos, along with roughly 240,000 years of audio content. The signal remains imperceptible to humans but survives light editing and compression, making it far harder for people to misrepresent AI-made content as human-created.

Easier access for everyone#

Previously, checking for SynthID required contacting Google's Gemini model directly—a clunky workaround that often produced unclear responses. Now, users can visit SynthID.com and upload files for analysis. Login is required through Google, OpenAI, or Apple accounts, and each user receives approximately 10 daily checks.

A key upgrade: the expanded detector now recognizes watermarks from all major participating AI firms, not just Google's own. This eliminates the need to visit separate detection tools for different platforms.

Current limitations#

The system does have meaningful blind spots. Major AI providers like Meta use different watermarking approaches, while others apply no watermarks at all. Open-source AI models that run locally can generate realistic content without any embedded markers. And despite Google's safeguards, someone may eventually develop techniques to strip SynthID from images, rendering them undetectable through this tool.

The website will carry a disclaimer acknowledging these constraints. Security researchers note that a more robust long-term solution might flip the approach entirely—using cryptographic standards like C2PA to certify authentic media at creation time rather than hunting for watermarks in fakes. Google's Pixel phones already support this alternative method, though adoption remains limited across the industry.