On 7 October 2026, Google officially launched its SynthID Detector website to the public, providing a centralized interface to verify whether images, video, and audio contain invisible AI watermarks (Times of India). Previously restricted to journalists and researchers following its initial release in May 2025, the verification tool handles approximately 1 million verification requests daily across its integrated ecosystems, including Gemini, Google Search, and Chrome.

Understanding SynthID Compatibility and Scope

Core Verification Mechanics

The web utility operates strictly by scanning for watermarks embedded by the SynthID system, which Google reports has been applied to more than 180 billion images and videos since 2023. At launch, the public detector is limited to identifying watermarks generated by specific partner models, including Google, OpenAI, NVIDIA, and Kakao (PCWorld). Support for Apple models is expected soon, though no fixed release date has been provided for that integration.

The platform accepts a broad range of file formats for analysis. Supported image inputs include JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF (TechCrunch). For video and audio, the tool processes MP4, MOV, WEBM, WAV, MP3, OGG, FLAC, AAC, and M4A formats (UA.News). Users navigating the public website face a daily usage quota capped at roughly 10 checks per user.

When a positive match occurs, the detector explicitly reports the specific partner model family responsible for the content (PCWorld). However, analysts emphasize that the mechanism is not a universal AI detector. A negative result merely indicates the absence of a detected watermark rather than confirming human origin (PCWorld), and the utility cannot recognize content generated by non-partner models or media created before watermarking was applied (The News).

While Google has expanded access through its public web portal, the underlying technology remains bound to participating developer ecosystems (PCWorld). Content generated by competitors outside the Google, OpenAI, NVIDIA, and Kakao alliance will bypass detection entirely (The News). Furthermore, media edited, compressed, or altered heavily enough to degrade the invisible signal can render the watermark unreadable to the verification tool (TechCrunch).

These operational boundaries underscore why industry watchdogs caution against treating the detector as a definitive solution for synthetic media verification (TechCrunch). Because the system relies entirely on proactive embedding at the point of generation, legacy AI files and outputs from closed-source platforms lacking SynthID integration escape identification (The News). Users examining ambiguous files must account for these technical blind spots when interpreting results (PCWorld).

Looking ahead, the next expected development for the platform involves integrating Apple model support into the public interface (9to5Google). Google has not yet announced a precise deployment timeline for Apple compatibility, nor has it detailed whether additional third-party artificial intelligence developers will join the watermarking partnership coalition (TechGenyz). Future updates will depend heavily on broader industry adoption of the SynthID protocol across competing software suites (Times of India).

Ultimately, Google's SynthID detector offers a scalable way to verify watermarked media but remains constrained by partner limits. Future updates will depend on expected Apple support and broader industry adoption, though current limitations mean it cannot serve as a universal detector (9to5Google, Times of India).