Google’s SynthID system has been used to debunk a high-profile AI-generated hoax image, in a rare but significant win for anti-deepfake technology.
Earlier this week, a picture circulated online that seemed to show Kentucky Senator Mitch McConnell covered in tubes in a hospital bed in a state of extreme distress. The image was shared widely on Reddit and X, but by Wednesday, the revered fact-checking site Snopes had debunked the image, noting that it registers as containing the SynthID watermark designed by Google to identify AI-generated pictures.
Senator McConnell’s health has been the subject of intense speculation since he checked into the hospital after an emergency call on June 14. Since that time, he has been largely absent from the public eye. In this case, however, the evidence clearly points to a fabricated image.
Launched at Google’s I/O developer conference in 2025, SynthID works as an invisible signature visible to SynthID algorithms but designed to be unnoticeable to the casual observer. Because the signature is built into the image itself, it survives even when an image is screencaptured across multiple platforms.
SynthID’s main limitation is that it can only be used when an image-generation tool actively participates in the program. Gemini models have included the watermark since the program launched in 2025. OpenAI joined in May 2026 as part of a broader effort to fight malicious image generation. Users can check if images contain the watermark by asking a Gemini model or uploading them to OpenAI’s public image verification tool.
