Google’s SynthID system was used to debunk a widely shared AI-generated image of Kentucky Senator Mitch McConnell appearing in a hospital bed. The image, which circulated on Reddit and X, was flagged by Snopes as containing a SynthID watermark, confirming its AI-generated origin. The system’s watermark, designed to be invisible to casual observers, proved effective in identifying the fake image. The incident highlights the growing role of anti-deepfake technology in countering misinformation.
The SynthID watermark, embedded into images during generation, remains intact even after screencapturing across platforms. This feature was critical in identifying the McConnell image, which was shared extensively online before being debunked. The watermark is visible only to SynthID algorithms, ensuring it does not interfere with the image’s appearance. Users can verify the presence of the watermark by uploading images to OpenAI’s public verification tool or querying a Gemini model.
The McConnell image emerged amid speculation about the senator’s health following his hospitalization on June 14. Despite his absence from public view, the image was debunked as entirely fabricated. SynthID, launched at Google’s I/O conference in 2025, is part of a broader effort to combat malicious image generation. While Gemini models have supported SynthID since its launch, OpenAI joined the initiative in May 2026, and Anthropic does not participate in the program.
Source: techcrunch
