Google Opens SynthID Detector to the Public Globally

Google DeepMind made its SynthID Detector tool available to the public globally at synthid.com on October 7, 2026, expanding a platform previously restricted to an early preview for researchers and journalists.

How the Detector Works and Supported Platforms

The free web portal checks uploaded files for imperceptible digital watermarks embedded during AI generation. Rather than analyzing visual artifacts to estimate whether content is synthetic, SynthID scans for mathematical patterns embedded in media pixels and sound frequencies.

To access the tool, users must sign in with a Google, OpenAI, or Apple account. The web interface accepts 20 file formats, comprising 11 image formats (including JPG, PNG, WEBP, and HEIC), three video formats (MP4, MOV, and WEBM), and six audio formats (including WAV, MP3, and FLAC).

The detector is not limited to Google models. It currently supports media generated using tools from Google, OpenAI, NVIDIA, and Kakao. Apple integration is scheduled to arrive soon, enabling the tool to identify assets generated or edited through Apple Intelligence features like Image Playground.

"As AI tools become better and easier to use, having clear context about the content you encounter online helps you make informed decisions about what to trust," said Pushmeet Kohli, VP of science and strategic initiatives at Google DeepMind. "Now, we're expanding access to everyone – with the tool available globally in English starting today."

Detection Boundaries and Limitations

The tool carries several functional constraints. Google's portal explicitly notes: "This is not a general AI detector. We can only detect media from the above companies, who have adopted SynthID technology." Content generated using independent models or platforms outside the partner ecosystem carries no watermark and will return negative results regardless of whether it was generated with AI.

The system also does not differentiate between entirely synthetic media and authentic files that underwent minor AI-assisted edits. Watermarked assets display a general status message, such as "This media was made or edited with Google AI," applying the same label to light retouching as to fully generated media.

Watermark Scale and Durability

Google initially launched SynthID in 2023. According to DeepMind, more than 180 billion images and videos and 240,000 years of audio have been watermarked with the technology. Google currently handles over 1 million verification requests daily across Google Search, Chrome, and the Gemini app.

Watermark resilience against intentional evasion remains a challenge. In independent testing, SynthID Detector identified 100% of intact original AI images, but the detection rate dropped to 18% after files were processed through a watermark removal tool. How reliably SynthID withstands heavy cropping, aggressive post-processing, and adversarial scrubbing in unconstrained web environments remains an ongoing uncertainty.