WASHINGTON — The video released by Democratic congressional hopeful Blake Gendebien last month looked, at first blush, eerily real: it depicted his potential rival, Representative Elise Stefanik of New York, dancing, posing with President Trump, and making a “shhhh” gesture with her index figure at her lips.
Except it wasn’t actually Stefanik doing any of those things. It was a version of Stefanik generated by artificial intelligence. A fine-print disclaimer at the end of the 45-second clip was the only explicit confirmation of how it was created.
The point of the video, Gendebien later told local press, was to satirically needle Stefanik for not-so-subtly maneuvering toward running for governor. Instead, the clip drew significant attention for its use of AI, with the Republican representative and her team slamming what she called a “deceitful” and “disgraceful AI digital ad” and local Democrats expressing discomfort with the strategy.
It was a demonstration of how generative AI, an increasingly powerful tool, could shape the 2026 elections like none other before it.
While the technology has been utilized in recent elections in the United States and abroad, political strategists and tech policy experts believe that as AI continues to improve — and grows more affordable and accessible — it will become increasingly ubiquitous in the political media landscape.
“It does feel like something is bubbling up,” said Hany Farid, a professor at the University of California Berkeley who studies AI content in politics. The increase, he said, can be explained by the fact that “the tech is just getting really, really good” and also by sheer political calculation.
“Once one side starts doing this, the other side is going to do it,” he said. “You’d be foolish to unilaterally disarm in this kind of politics.”
As the midterm election cycle ramps up, candidates, major national party committees, and super PACs across the political spectrum are using AI to create videos and images, often to mock or attack their opponents.
The National Republican Senatorial Committee, for instance, recently posted to X a realistic AI-generated image showing Senator Adam Schiff of California and Senator Chris Van Hollen of Maryland sitting and drinking margaritas together, invoking GOP mockery of Van Hollen’s trip to El Salvador to check on his detained constituent Kilmar Abrego Garcia.
And the Democratic Congressional Campaign Committee posted on X an AI-generated image of Trump, looking frightened and cartoonishly orange, with the caption “TACO,” which stands for the taunt that “Trump always chickens out” on tariff threats.
Currently, there is no federal policy to regulate the use of AI in political communication. In 2023, Public Citizen petitioned the Federal Election Commission to adopt new rules explicitly establishing that “deepfakes” fall under existing law prohibiting federal candidates from “fraudulently misrepresenting” other candidates or political parties. But the FEC did not take it up.
However, 26 states have enacted some type of law designed to regulate AI in politics, from requiring disclosures on AI-generated ads to banning the publication of deepfakes close to an election. (Gendebien had a disclaimer on his video, while the NRSC and DCCC do not use disclaimers on their AI content.)
The types of deepfakes that have most alarmed policymakers are images or videos designed to convince voters that someone did or said something they did not — a practice that has already disrupted campaigns in the United States and elsewhere. Using AI to create exaggerated animations — depicting political figures posing or looking a particular way — has proven less controversial.
But to experts like Farid, it’s the more mundane content replacing traditional photos and animations that most captures what the new AI reality of politics will actually look like.
“It’s propaganda,” he said. “The thing with propaganda is it’s been around forever, but think about how easy it is to make now, how much more powerful.”
Many tech policymakers and watchdogs are concerned about where the trend is heading, like Ilana Beller of the nonpartisan advocacy group Public Citizen, which has worked with members of Congress on legislation to regulate the use of AI in political content.
“Normalizing the use of AI in our media ecosystem broadly, but certainly in our political information system, absolutely carries the risk of undermining trust and belief and the trust that is so critical in our information and our media ecosystem,” Beller said.
Mark Alfano, a philosophy professor at Macquarie University in Australia who studies AI, has coined a different term for the phenomenon: “slopaganda.” The word is a portmanteau of propaganda and “AI slop,” a term used to describe quickly constructed and often low-quality AI content.
“It does seem to be really popular,” said Alfano, who noted an increasing number of examples of the White House pushing out AI content. In July, Trump’s team posted on X an animation of the White House covered in gold with gold coins and papers falling from the sky, set to Bruno Mars’s “24K Magic.” (The post mocked “haters” who might say “tHiS iS AI!!”)
The 2024 election cycle was the first to see widespread use of AI, which often proved controversial. During the Republican presidential primary, Florida Governor Ron DeSantis circulated realistic-looking AI images of Trump hugging Dr. Anthony Fauci. Then-senator JD Vance responded that “smearing Donald Trump with fake AI images is completely unacceptable.” Later, Trump’s campaign used AI to generate images depicting his opponent, Kamala Harris, addressing an arena full of communists and adorned with communist symbols.
Generative AI tools are now even faster, more powerful, and more widely available than even last year. They are particularly suitable to the fast-paced world of political communication, said Tyson Brody, a veteran national Democratic strategist.
“The one thing you don’t get more of in a campaign is time, and the fact you can save time is just gonna lead to more” AI, Brody said. “Campaigns are going to do it, people are going to do it. It’s going to blur the lines a lot more than they were before — it’s just a continuation of things in the past but it’s going to be more powerful and more out there.”
While many experts agree that disclaimers are better than nothing, there’s shared concern over the fact that little is truly known about how voters process AI content in political media.
Beller, of Public Citizen, said many “people go on social media and don’t realize what’s real.”
Alfano, the professor, pointed to one image in particular that shows how AI images can cloud perceptions. When Hurricane Helene devastated the Southeast shortly before the 2024 election, an image went viral of a crying young girl clutching a puppy in a rescue boat.
Many right-wing figures shared the photo, hoping to prove that the Biden administration’s response to the hurricane was failing. When it was revealed to be an AI fake, some Republicans continued to promote the image, saying it represented how people were feeling about the disaster.
It’s a troubling standard, said Alfano.
“It’s not that it’s true, it’s not that it provides evidence, it’s not that it’s reliable, it’s not that it’s coming from a trusted source,” he said. “It’s just emblematic of how people feel. And there’s almost no way to argue against that kind of thing.”
Sam Brodey can be reached at sam.brodey@globe.com. Follow him @sambrodey.