AI election parodies spread after midterms as one clip tops 16 million views
A New York Times report finds absurd political videos reaching large audiences. Experts caution that obvious fakery can still rally supporters and sharpen attacks.
Listen to this story
The audio brief
Story brief
3 key pointsFollowing the 2026 U.S. midterms, the most visible AI election videos were often unmistakable jokes rather than fake footage presented as real; a digitally recreated Office cast clip drew more than 16 million views on X. That public pattern does not settle the deception risk: research cited by The New York Times found hundreds of AI campaign ads, including deceptive ads that did not disclose AI use, while researchers warn realistic fakes may circulate in private partisan groups. Viral reach and misleading impact are different measures, and public memes alone cannot show whether fakes changed votes.
- 01
Experts cited by The New York Times found no deceptive deepfake of James Talarico presented as genuine, despite many generated videos of him dancing and singing.
- 02
An AI ad depicted Vikki Goodwin endorsing a so-called “death tax”; she said unfamiliar viewers might mistake it for a real position, and she won by more than six points.
- 03
Research on Canada’s previous federal election found AI material in about 6% of election-related images on X, while more than 90% of views went to comparatively innocuous content.
After Democrats swept the 2026 midterms, AI-generated celebrations turned election winners and television characters into viral performers. One parody featuring replicas of The Office cast drew more than 16 million views in a single X post, according to a November 12 New York Times report—a striking example of political fakery built for laughs.
The results came first, then the synthetic celebrations
The Office skit imagined its boss adapting to a progressive society. He wore a Pride flag sweater and called for an emergency Diversity Day. The premise was an invented celebration, not footage of an actual workplace gathering. Its audience on X made it a prominent example of the post-election wave described by Times reporter Stuart A. Thompson.
Other clips put incoming senators into similarly fabricated scenes. An AI version of Texas Senator-elect James Talarico pushed aside his Republican opponent, Ken Paxton. Another showed Michigan Senator-elect Abdul El-Sayed dancing with New York City Mayor Zohran Mamdani. The Times credited those examples to TBC_on_X.
These were not isolated uses of comic exaggeration. Thompson described many of the election season’s most popular AI videos as outlandish jokes and parodies, showing events that obviously never happened. That pattern contrasted with the long-running fear that convincing digital impersonations would mislead voters about invented scandals or statements.
Obvious fiction can still carry a political message
Talarico’s campaign offers a useful distinction. Although ample material existed to create fake video or cloned audio of him, experts cited by the Times could not find a deceptive deepfake presenting itself as genuine footage. Instead, dozens of generated videos showed him dancing and singing, collecting millions of views.
The jokes were not all friendly. Opponents of El-Sayed circulated videos disparaging him, often using racist stereotypes. Other memes mocked President Trump as a small-business owner, with Vice President JD Vance cast as his son and employee. The same medium served celebration, ridicule and attacks.
Henry Ajder, an AI expert and founder of Latent Space Advisory, told the Times that such videos can rally supporters and make criticisms more memorable. He compared their political purpose with familiar campaign tactics, while distinguishing the new tools’ scale, accessibility and emotional force.
The case that everyone’s worried about is a highly inflammatory or incriminating deepfake going viral on the eve of an election and truly moving the needle on how people have voted.
Henry Ajder, speaking to The New York Times
Parody’s reach does not settle the deception question
Ajder said that feared election-eve scenario had not yet materialized. But the season also included more realistic impersonations. The Times cited Wesleyan Media Project research identifying hundreds of AI-generated campaign ads, including some that were deceptive and did not disclose their use of AI. Most depicted obviously impossible scenes.
One more grounded ad depicted an AI version of Vikki Goodwin, then a Democratic candidate for Texas lieutenant governor, purportedly supporting a “death tax.” Goodwin told the Times in September that people unfamiliar with her might mistake it for her. The campaign behind the ad called it an obvious parody produced for entertainment. She ultimately won by more than six percentage points.
The attention pattern has a parallel outside the United States. Research from McGill University and the University of Montreal, cited by the Times, found AI-generated material in about 6 percent of election-related images on X during Canada’s previous federal election. More than 90 percent of views went to comparatively innocuous material, including memes and portrayals of candidates as superheroes or supervillains.
Public viral hits also leave a harder-to-see question. Researchers told the Times that deceptive fakes may spread more easily through group chats or private groups, targeting partisans inclined to believe them rather than more skeptical swing voters. The visible meme party is therefore only one part of the election’s synthetic-media story.
Sources
- nytimes.comWe Were Warned of Deceptive A.I. Fakes. We Got Memes.
Reader comments
Newest comments first. Replies stay oldest first.