AI-generated images from 2026 political ads targeting James Talarico, Mike Mazzei, Peggy Flanagan, Kathy Hochul and Zohran Mamdani. (NRSC; Make Oklahoma Great Again PAC via Oklahoma PBS; North Star Dawn PAC via KSTP; Blakeman for Governor)
The midterm elections are around the corner, and campaigns and PACs have already spent more than $137 million to air AI-made ads, including an 85-second video in which Texas Senate candidate James Talarico looks into the camera, reads his old tweets aloud, and voices spicy commentary he never actually made. The Senate Republicans’ campaign arm produced it. UC Berkeley deepfake expert Hany Farid said most people wouldn’t immediately know it was fake; the “AI GENERATED” tag sits small and faint in a corner.
Between now and Nov. 3, any clip of a candidate on your TV or in your feed may be fabricated, and most carry no AI label whatsoever. I’ve been predicting a deluge of deepfakes, and fear that it will get worse if we don’t act.
James Talarico, in his own words: pic.twitter.com/lDlUoqBbP7
— Senate Republicans (@NRSC) March 11, 2026
In 2024, I donated a year of my life to fighting political deepfakes by founding the nonprofit TrueMedia.org. A talented team at TrueMedia.org developed a state-of-the-art deepfake detector in record time and distributed it broadly, free of charge. It analyzed more than 60,000 images, videos and audio clips in elections worldwide, but it was early.
The Wesleyan Media Project, which reported the $137 million figure, found less than $50,000 in AI ad spending for the entire 2024 cycle. In January 2025, I said , “It did not materialize in 2024. That doesn’t mean that it’s not coming.” We open-sourced the technology and transferred it to Georgetown University, which resurrected TrueMedia.org in time for the peak of the current election cycle.
I was expecting foreign operatives to be the most prolific fakers. But in 2026, the deepfakes are coming from American campaigns and PACs, on television and in your feeds.
Many of the AI ads the Wesleyan Media Project has cataloged are harmless. Some are cosmetic: a still photo of Oklahoma candidate Chip Keating animated into a smile, or a synthetic family at a kitchen table where a campaign used to buy stock footage. Fred Davis, who has made campaign ads for 50 years, told The New York Times a scene that would have cost $2 million to shoot now costs about $200,000 to generate. Nobody is deceived.
Others are absurd on their face: New York’s governor and mayor gardening together while pitching a “political delusion pill” in a Bruce Blakeman campaign ad, or Spencer Pratt as Batman leading fed-up Angelenos armed with tomatoes. Pratt finished third in the Los Angeles mayoral primary. These are political cartoons with a pulse, and they’re protected satire. NYU’s Scott Babwah Brennen calls the genre “AI-generated slopaganda”: political slop made to stir feelings rather than make an argument.
. @KathyHochul and @ZohranKMamdani think they can cure New York’s cost-of-living crisis with excuses, empty promises, and “delusion pills.”
The real side effects of their policies aren’t a joke: 48 utility rate hikes, higher taxes, uncontrollable government spending, and… pic.twitter.com/CWxttSRwUK
— Bruce Blakeman (@NassauExec) September 1, 2026
The third kind is built to be believed, like the Talarico video. A Make Oklahoma Great Again PAC ad showed gubernatorial candidate Mike Mazzei cheek to cheek with Hillary Clinton under the words “Clinton Lover. Trump Hater,” with no AI label; the PAC spent more than $200,000 putting it on Oklahoma television. A PAC ad in Minnesota showed Senate candidate Peggy Flanagan with her hand out in front of a pile of cash. Talarico’s own ad now ends with the words “Made by humans, not A.I.”
A North Star Dawn PAC ad showed Minnesota Senate candidate Peggy Flanagan in front of a pile of cash. (North Star Dawn PAC via KSTP)
A deceptive fake doesn’t need to be high quality. In a 2021 experiment with 5,750 people, a deepfake of Elizabeth Warren convinced 47% that a fabricated scandal was real; a plain fake headline convinced 43%.
A Utah Valley University study this June found voters identified deepfakes correctly about 16% of the time, and people who called themselves “extremely familiar” with deepfakes did no better than people who said they weren’t familiar at all. Media literacy doesn’t inoculate you against deepfakes.
The laws on the books fail to solve the problem. At least 31 states have deepfake election statutes. Wesleyan’s researchers found 29% of AI ads labeled in states with laws and 32% in states without.
Even when a label appears, voters often miss it. NYU’s Center on Technology Policy showed 1,051 people labeled and unlabeled ads. Thirty-seven percent didn’t remember seeing Michigan’s required disclaimer at all, and because it says “manipulated by technical means” rather than “AI,” only 21% identified AI as the technology involved. The one thing labels reliably did was make viewers trust the candidate who ran the ad less, which is why campaigns leave them off.
My proposed solution is simple: the litmus test should be the person shown, not the technology used. Did a real person agree to be shown this way? If not, the ad carries a plain-language label, on screen and in the audio, or it doesn’t air. A station manager can apply the test in a minute, which addresses the complaint that nobody can define “deceptive.” The rule should cover paid political ads, the one place a broadcaster already checks every spot before it airs; memes, satire accounts and your cousin’s group chat are a different problem.
Sejin Paik, the Georgetown researcher who now leads TrueMedia, puts the television problem plainly: “On TV, where AI ad spending is rising, a viewer watching a deceptive ad has little room for a crowd reaction or a provenance mark to change their view.”
H.R. 9922 would set a federal disclaimer rule, and the consultants’ own trade group is asking for a federal standard to replace the state patchwork. When the regulated party asks for the regulation, pass it.
Voters are ahead of Congress: in a Scripps News poll this month, 43% of registered voters said they had already been fooled by an AI-made ad this campaign, and 77% of Democrats and 79% of Republicans want rules on AI in politics.
Looking to the future, the 2028 election is a presidential one — we don’t want to be flooded with even more deceptive AI ads. Two years is enough time to get the rules and processes in place, if we start now.
