Google has made its SynthID Detector available to the public, allowing anyone to upload an image, video, or audio file and check for the invisible watermark that Google's AI models embed. The tool previously was limited to researchers and journalists.
The detector scans for SynthID watermarks rather than analyzing content for signs of AI generation. That approach avoids false positives but means it only works with AI models that use SynthID, including Google's Gemini and models from OpenAI, NVIDIA, and Kakao. Apple participates but support isn't ready. Meta's tools are not covered.
Google limits users to about 10 checks per day and requires login with a Google, OpenAI, or Apple account, reportedly to prevent people from using the detector to reverse-engineer and remove the watermark. The company says watermarks persist after editing, though no guarantee against future circumvention.
For businesses, the tool is a practical but partial solution. Other detectors exist, but they are also tied to specific markers. General-purpose AI detectors remain unreliable and prone to false positives. The source notes that no reliable cross-platform detector is publicly available.