Google has released the new SynthID Detector website, allowing anyone to check images, videos, and audio files for hidden generative AI watermarks.
As generative AI tools continue to flood the internet with synthetic images, videos, and music, determining whether digital media is authentic has become a daily challenge. According to Engadget, Google is expanding public access to its detection technology by launching a dedicated web tool at SynthID.com. The system allows regular users to upload files and receive a straightforward yes or no answer regarding whether a watermark is present, a notable step up from a limited rollout given to researchers and journalists in 2025.
Google originally introduced SynthID in 2023 to embed invisible labels directly into content generated by its Gemini model. While the company previously required users to ask Gemini itself to verify a file, the new website provides a dedicated verification channel. Beyond Gemini, the detector can also scan for watermarks embedded by SynthID partners, which include OpenAI, NVIDIA, and Kakao. Apple also participates in the initiative, though detector support for Apple creations is still in development according to the report.
Because the detector relies entirely on scanning for specific watermarks rather than analyzing visual patterns, Engadget notes that it avoids false positives. However, this simplicity brings significant constraints. The tool cannot detect media produced by AI models that do not use SynthID, such as Meta systems, nor can it pinpoint specific portions of a file that were artificially altered. To protect the technology from exploitation, Google restricts users to roughly 10 checks per day and requires a login through a Google, OpenAI, or Apple account.
Background
Google initially introduced the SynthID framework in 2023 to apply imperceptible labels to AI-generated content at the point of creation. The company maintains that these signatures persist even if the underlying media files are edited later. While other technology firms have deployed their own watermarking strategies, such as Meta using a marker called Content Seal, universal detection tools remain rare. Many researchers have built custom detectors for private use, while publicly available general purpose detectors often struggle with accuracy and frequently mislabel human-created work as artificial.
Why It Matters
The public rollout of the SynthID Detector highlights the growing industry shift toward collaborative verification standards rather than relying on guesswork or unreliable pattern analysis. By allowing cross-company detection for partners like OpenAI and NVIDIA, the tool points toward a future where major tech firms might establish shared accountability infrastructure. At the same time, the fragmented nature of current detection tools means users cannot rely on a single solution to verify all internet media. As generative media becomes increasingly sophisticated, the limits of watermark-dependent systems suggest that media literacy and multi-layered verification will remain vital for digital trust.
Illustration: AI-generated image, not a photograph of the event.
Source