AI political ads warp reality ahead of US midterms

Washington: A political ad depicts a US Senate hopeful remarking on God and gender while scrawling in a Bible with a crayon. But he never filmed the clip: it is yet another AI-generated video pushed by a rival ahead of the midterm elections.

AI-created ads -- placing politicians in fictitious situations -- have proliferated ahead of the November elections, a largely unregulated form of political communication that is increasingly being adopted despite a broader backlash against the technology.

Earlier this month, Republican Senate nominee in Texas, Ken Paxton, posted an AI-generated ad on X depicting his Democratic rival James Talarico saying: "Look, I always think of myself as a Christian who hates Christianity."


The lifelike clip shows Talarico writing "let there be six genders" in a Bible with multicolored crayons. Overlaid text reads: "James Talarico wrong for Texas" and "wrong for America," while a QR code seeks donations for Paxton.

Talarico condemned the ad, saying "they are lying about a candidate in this race" and that he favored a ban on deepfakes in political campaigns.

"These AI deepfake ads don't add anything to our civic discourse," he told NBC News. "They don't do anything to elucidate the choice that's in front of voters."

The ad came a few weeks after a conservative nonprofit aligned with President Donald Trump released another AI clip depicting Talarico parodying Julie Andrews's character in the hit classic Sound of Music, dressed in a jumper-style bodice and skirt and singing "My favorite things."

'Erodes trust'

The proliferation of AI ads in US federal, state and local elections comes against the backdrop of declining trust in the technology.

"The concern might be not that people believe the ads but the ads and general AI slop content risk people doubting everything," Sarah Kreps, director of the Tech Policy Institute at Cornell University, told AFP.

"People might just dismiss real content as another deepfake, which erodes trust, causing voters to tune out rather than dig deeper for authenticity."

The Wesleyan Media Project, which tracks election ad spending, has identified at least 164 political ads nationwide so far this year that were "generated or enhanced through the use of AI." It estimates that the ads cost just under $80 million.

That tally of ads -- sponsored by both Republicans and Democrats and splashed across airwaves and social media -- was an undercount.

The group said it was difficult to definitively know whether AI tools had been used in some ads, especially as there were no federal mandates that political ads carry an AI disclaimer.

"Just the knowledge that the tech has become good enough to deceive the average viewer combined with the fact that it is being used in election communication is enough to undermine voters trust in the broader election communication ecosystem," Ilana Beller, from the nonprofit advocacy group Public Citizen, told AFP.

'Modern political cartoon'

However, cheap and widely available AI tools help save on the often exorbitant costs of producing ads.

"There's no need to pay for a camera crew or hire actors to create good-looking content on a short timeline. They also open the door to creativity -- no special effects editors required," Isabel Linzer, from the nonpartisan Center for Democracy & Technology, told AFP.

In many cases, Linzer said, satirical or humorous AI ads serve as "the modern political cartoon."

"A worthy question is whether using AI in political ads is a stunt that will boost a candidate through generating substantial free media attention," she added.

The AI-generated ads could help candidates with goals such as fundraising and mobilizing supporters, but there is scant evidence that they influence how people vote, said Scott Babwah Brennen, director of the Center on Technology Policy at New York University.

"Much of this does seem to be effective in gaining attention, though it is unclear if that will translate into electoral success," Babwah Brennen told AFP.

"As public sentiment grows against AI, it is possible that AI-generated negative attack ads could backfire -- that is decreasing assessments of the producer of the ad."