Candidates and their allies have spent more than $80 million this election cycle on political ads made with artificial intelligence, according to the Wesleyan Media Project, and most of it came from Republicans ahead of the Nov. 3 midterms.

The project counted more than 164 AI-based ads, with about 80 percent paid for by Republicans and their backers, Al Jazeera reported on Friday. Thirty-one states require such ads to carry a label. There is no federal law.

Bruce Blakeman, the Republican nominee for governor of New York, has run ads in which AI versions of Gov. Kathy Hochul and Mayor Zohran Mamdani of New York City garden together and work out how to raise prices on residents. In another, a fabricated Ms. Hochul talks about raising tolls and fares so people can’t afford to drive or ride the subway. Neither of them appeared in the ads. One of the videos runs 43 seconds and opens with the two strolling through a park on a sunny day and introducing themselves, 404 Media reported.

Mr. Blakeman said his ads carry AI disclaimers and belong to an old tradition of political satire. “People find it humorous. They like satire,” he said.

The labeling is in dispute. Al Jazeera said the disclaimer appears for a few seconds at the start of the ads. The Associated Press, looking at the park video, found only small text reading “paid for by Blakeman for New York” and no mention of AI, though New York law requires that disclosure. The outlets may have been looking at different versions. In August, the New York State Democratic Party filed a complaint with the state Board of Elections saying Mr. Blakeman broke a recently amended election law that requires materially deceptive political material to be labeled as AI-generated. It was not clear whether the board had ruled.

New York isn’t the only state where this is happening. AI ads have cast Sherrod Brown, the Ohio Democrat, as a party guest who won’t leave, put James Talarico, a Texas Democrat, in a blue dress styled on The Sound of Music and turned Mike Rogers, a Michigan Republican, into a superhero.

Scott Babwah Brennen of New York University’s Center on Technology Policy said this season belongs to “slopaganda,” ads that are plainly fake. They try to “communicate deeper truths rather than factual truths,” he said, and could “impact voter trust.”

Democrats have tried it too. In San Francisco, State Senator Scott Wiener, a Democrat running to succeed Nancy Pelosi in the House, released an AI chatbot that answered about 5,000 questions while mocking his opponent, Connie Chan, over her record on development, her accent and her status as a naturalized citizen. He took it down after people called it bigoted, saying “it missed the mark.” Mr. Wiener wrote a California bill on AI safety disclosures, and a weaker version became law. His office did not respond to Al Jazeera’s request for comment.

Adam Rose of Starling Lab, who opposes banning AI ads and says minor edits like color correction shouldn’t need a label, said the chatbot “showed bad judgement. It showed voters something about him.”

Ian Krager, a spokesman for Ms. Chan, called the race “a harbinger for the country” and said she stands with working families while Mr. Wiener sides with AI companies.

Ilana Beller of Public Citizen said some candidates have pledged not to use AI ads, and voters seem to like them for it. She wants AI labels to be as clear as the ones on food. “There is still a need for a federal law,” Ms. Beller said. “People need to know to look for labels.”