Key facts
- Google's SynthID system was used to debunk an AI-generated hoax image of Senator Mitch McConnell.
- The image, showing McConnell in a hospital bed with tubes, circulated widely online.
- Snopes confirmed the image contained a SynthID watermark, indicating it was AI-generated.
- SynthID embeds an invisible signature into images, which survives even when images are screencaptured.
- Gemini models and OpenAI's tools include the SynthID watermark, while Anthropic does not.
Google's SynthID system has successfully debunked a high-profile AI-generated hoax image depicting Senator Mitch McConnell in a distressed state in a hospital bed. The image circulated widely on social media platforms like Reddit and X before the fact-checking organization Snopes identified it as fake. Snopes noted that the image contained a SynthID watermark, a feature designed by Google to invisibly embed a signature into AI-generated pictures, making them identifiable. This marks a significant success for anti-deepfake technology.
Senator McConnell's health has been a subject of public speculation since he was hospitalized on June 14. The AI-generated image fueled these concerns, but evidence proved it to be entirely fabricated. Launched at Google's I/O developer conference in 2025, SynthID works by embedding a signature that is invisible to the human eye but detectable by SynthID algorithms. This signature persists even when images are screencaptured across different platforms.
A key limitation of SynthID is its reliance on the participation of image-generation tools. Google's Gemini models have included the watermark since the program's inception in 2025. OpenAI joined the initiative in May 2026 as part of a broader effort to combat malicious image generation. Anthropic, however, does not currently participate in the program. Users can verify images by using a Gemini model or OpenAI's public image verification tool.