Google's SynthID watermarking system has been used to expose a high-profile fake image, marking a notable success for anti-deepfake technology. An AI-generated picture purporting to show US Senator Mitch McConnell severely ill in a hospital bed spread widely on Reddit and X earlier this week, but the fact-checking site Snopes debunked it after finding the image carried Google's SynthID watermark, which identifies AI-generated content. The case matters because it demonstrates the tool working as intended amid growing concern over convincing political deepfakes.

The hoax played on genuine public speculation about McConnell's health, which has been intense since he was admitted to hospital following an emergency call on 14 June and has remained largely out of view. SynthID, launched at Google's I/O conference in 2025, embeds an invisible signature that survives screenshots across platforms but is only present when an image-generation tool participates in the scheme. Google's Gemini models have carried it since launch and OpenAI joined in May 2026, though Anthropic does not take part. Users can check images via a Gemini model or OpenAI's public verification tool.