Google Launches Public SynthID Detector for AI Content, but Limits Remain

Users must sign in with a Google, OpenAI or Apple account to use the detector. Google said the tool will be expanded in a few weeks to identify images created with Apple Inc.’s AI tools, such as Image Playground. Google first developed SynthID in 2023 as an imperceptible watermark added to AI-generated content from the participating companies. Detecting that watermark previously required special tools, and Google had made versions available only to journalists and other media professionals. The company also offers other ways to check for AI content, including its Circle to Search smartphone feature and the option to upload content to Gemini and ask whether it was created with AI. The dedicated detector makes those capabilities more accessible.

The main limitation is that SynthID Detector is built to find SynthID watermarks, so it is not effective at detecting AI-generated content that does not carry one. Anthropic PBC’s Claude does not use SynthID, nor does xAI Corp.’s Grok chatbot. None of the many open-weights Chinese AI models use SynthID either, meaning the detector will miss a large amount of AI-generated material. The tool also cannot distinguish between media created entirely from scratch and media edited with AI, and it cannot show which part of an edited item is fake and which is real.

The launch comes amid a broader absence of foolproof verification for AI-generated content, partly because watermarking standards differ across model makers. Meta Platforms Inc. has been experimenting with its own watermarking tools and has developed a system for identifying images created with Muse Image, but SiliconANGLE reported that system is useless against images created with third-party models. Unless all major AI model makers, including Chinese companies, adopt a universal watermarking standard, reliable detection of AI content is unlikely, according to the report.

Google appears to recognize the problem. The company said on the SynthID Detector website that it is “continuously working on extending our partnerships” so it can detect more types of AI-generated media. “Building industry-wide standards takes time, and different content types (image, audio, video) have different technical requirements,” Google said. “The goal is a layered ecosystem where watermarking, metadata standards, and verification tools work together across the industry.” SiliconANGLE reported that Google, given its influence and existing partnerships with several technology heavyweights, is likely in the best position to help create a universal standard.