Google's SynthID debunks fake Mitch McConnell hospital photo

An image circulated online this week that seemed to show Kentucky Senator Mitch McConnell covered in tubes in a hospital bed in a state of extreme distress. The picture spread widely on Reddit and X before Snopes, the fact-checking website, debunked it on Wednesday after finding that it contained Google’s SynthID watermark — an invisible signature designed to identify AI-generated images.

The hoax landed amid intense speculation over McConnell’s health, which has run high since he checked into the hospital after an emergency call on June 14. His absence from public view since then has fueled speculation that his health may be failing. The SynthID check showed the viral image was fabricated rather than evidence of the senator’s condition.

SynthID launched at Google’s I/O developer conference in 2025, embedding a signature into images that is invisible to viewers but detectable by SynthID algorithms. Because the signature is built into the image itself, it survives even when an image is screencaptured across multiple platforms, as happened when the McConnell picture moved from its point of origin onto Reddit and X.

The system’s limitation is that it only works when the tool that generated an image actively participates in the program. Google’s Gemini models have included the SynthID watermark since the program started in 2025, and OpenAI joined in May 2026. Anthropic does not participate, meaning images generated by its models carry no equivalent watermark for Snopes or anyone else to check.

Why it matters

The McConnell case gives Google’s SynthID a concrete real-world result: proof that its invisible watermark can survive being shared and screenshotted across different platforms and still be used to correctly identify an AI-generated image. It’s a rare validation for a class of anti-deepfake tools whose effectiveness has mostly been theoretical until now.