Callista Gingrich, the U.S. ambassador to Switzerland and Liechtenstein, honored Swiss National Day with a video that went viral for all the wrong reasons on Aug. 1. It launched a slew of online reactions ranging from CNN’s Andrew Kaczynski saying he felt like he was “dissociating” to Civitas Institute Director Nick Lindquist poking at “some Gen Zer working for her that thinks it’s hilarious to have her do this.”

Viewers, and even the experts HuffPost consulted, are now speculating whether this video and others like it are AI-generated.

Siwei Lyu, a professor in the Department of Computer Science and Engineering at the University of Buffalo, runs the Deepfake-o-Meter lab that compares videos, photos and audio to AI media-detection models. Lyu told HuffPost he suspected the video could be AI because of Gingrich’s eyes.

“One telltale sign is the highlights in her eyes,” Lyu said. “They should correspond to static light sources in the real environment, but in the video the highlights are moving with the eyes with no relative motion.”

There are two light reflections in Gingrich’s eyes that can be seen on either side of her pupil in the Aug. 1 video, as well as her earlier July 4 and June 11 videos.

At HuffPost’s request, Lyu’s lab ran the Aug. 1 video through the Deepfake-o-Meter, which tests footage against several AI-detection algorithms and, in this case, compared it to footage of Gingrich speaking at public events. The video resulted in a 99.9% score in the lip-sync-aware models and low scores on the frame-level detectors, which is expected for a deepfake of a real person, according to the lab’s report.

Based on those results and manual findings, the report claims the video is AI-generated.

The lab’s findings noted several inconsistencies: Gingrich’s neck anatomy didn’t move the way it should for her head movements; her necklace appeared to subtly change shape between frames rather than rotate naturally; her teeth looked inconsistent from frame to frame; and the lip-sync detector returned a score consistent with those typically seen in AI-generated avatars or lip-sync deepfakes.

The U.S. Embassy in Switzerland and Liechtenstein did not respond to HuffPost’s request for comment as of publishing.

“The tipoff is that her voice is too smooth,” Traci Brown, a body language expert and behavior analyst, told HuffPost. Brown also noted that Gingrich doesn’t seem to blink enough or swallow while speaking. Gingrich blinks five times in the 28-second Aug. 1 video, which roughly works out to 10 to 11 blinks per minute, well below the standard range.

“Normal blink rate is 15-22 blinks per minute,” Brown said.

Not every expert HuffPost spoke to was convinced. This is the seventh video Gingrich has been in since she was confirmed as the Ambassador in September 2025. All of them are similar: Gingrich speaks directly into the camera, most of the time in front of a green screen, with a slight twist to her right, which body language deception expert Susan Constantine told HuffPost is traditionally the best angle on camera.

“We’ve gotten normalized to second-guessing online videos like this — straight to the camera, with beauty filters — because of the proliferation of fake avatars,” Sam Gregory, the executive director of global human rights organization WITNESS and its Deepfake Rapid Response Force, told HuffPost.

“[But] the body not moving is not necessarily a sign of AI, just a sign of an uncomfortably posed person,” he said. “The reality is that the visual and audio clues we’re told to look for [in AI] are disappearing.”

Gingrich’s tone and voice are also “very scripted,” Constantine said. “It’s very possible she’s probably rehearsed this many times until she got it right.”

A big driver for people to question whether something is AI-generated is authenticity, Constantine said. Even for politicians, years of social media have trained viewers to expect some baseline of authenticity in speeches and videos. Constantine noted that President Donald Trump particularly emphasizes this expectation by injecting his personality when he speaks in public or on camera.

“If she was not performing and she was genuinely expressing and had a bullet point or something to read off from — rather than scripting it word for word — it would have been a whole different scenario,” Constantine said. “When you’re too robotic, and your movements and everything are all within a rhythm — you’re trying to frame things up and everything. People love it when you goof up.”

By entering your email and clicking Sign Up, you're agreeing to let us send you customized marketing messages about us and our advertising partners. You are also agreeing to our Terms of Service and Privacy Policy.