Republican candidates, party organisations and their allies accounted for approximately $169 million (£125.5 million) in spending on AI-generated and AI-enhanced political advertisements during the 2025–2026 US election cycle, according to research published by the Wesleyan Media Project.
The figure represented just over 90% of the estimated $187 million (£138.8 million) spent on such advertising, as Democrats secured major victories in the November 2026 midterms.
The scale of the spending marks a sharp increase from the previous election cycle.
The Wesleyan Media Project, which tracks political advertising, identified less than $50,000 (£37,100) in spending on AI-generated political advertisements during the 2024 election cycle. Its latest research identified 289 unique advertisements generated or enhanced using AI, although the group warned that its figures were likely an undercount.
AI Election Ads Became a Republican Campaign Tool
Republican entities sponsored 82% of the 289 advertisements identified by the Wesleyan Media Project, compared with 13% sponsored by Democratic campaigns. Independent candidates and groups accounted for the remainder.
The spending breakdown was even more heavily weighted towards Republicans. Candidates, a Republican party committee and Republican-leaning groups were responsible for just over 90% of the estimated total, or approximately $169 million (£125.5 million).
Some of the advertisements relied on fabricated scenes involving prominent political figures.
In Oregon, Republican congressional candidate Jonathan Lockwood used AI-generated videos during his unsuccessful campaign for the party's nomination in the 5th Congressional District.
One video depicted Democratic Representative Janelle Bynum and Governor Tina Kotek in an apparent Satanic ritual involving frightened children, according to Oregon Public Broadcasting.
Lockwood's campaign accompanied the material with the caption, 'No one's sons or daughters are safe under Tina Kotek and Janelle Bynum.'
The imagery was fictional. Bynum was the Democratic incumbent Lockwood hoped to challenge in the general election, while Kotek was seeking another term as governor. Lockwood lost the Republican primary to Patti Adair on 19 May.
The advertisements also attracted scrutiny beyond the campaign trail.
Oregon's Secretary of State's Office investigated whether Lockwood had violated the state's AI-disclosure requirements. The investigation did not establish that he had broken the law.
Lockwood disputed the allegation and said he would address it through the legal process.
AI Ads Targeted Candidates With Satire and Fictional Scenes
Other campaigns used AI to produce more theatrical attacks on their opponents.
Filmmaker Charles Curran posted an AI-generated promotional video supporting former reality television personality Spencer Pratt's campaign for Los Angeles mayor.
The Batman-inspired clip portrayed Pratt as a superhero fighting fictional versions of political figures, including California Governor Gavin Newsom, incumbent Mayor Karen Bass and Vice President Kamala Harris.
The video reportedly attracted more than five million views. Pratt reposted it on X, although his campaign said it had not commissioned the advertisement.
In North Carolina, an AI-generated re-election advertisement for Republican state Senator Mike Lee portrayed his Democratic opponent, Jessica Bichler, as a caricature of liberalism. The campaign video used the nickname 'Bonkers Bichler' and depicted her in an exaggerated cartoon-like form.
Bichler defeated Lee in the 3 November election, receiving 52.64% of the vote against his 47.36%, according to unofficial state election results reported by Carolina Journal.
AI-Generated Ads Leave Their Actual Influence Unclear
The election results delivered a major victory for Democrats, who won control of both chambers of Congress.
However, the outcome does not establish that AI-generated advertisements contributed to Republican defeats or that voters rejected candidates because of their use of the technology.
There are also signs of growing resistance to AI infrastructure, although that is a separate issue from the political advertising figures.
Local voters backed more than 20 measures restricting or opposing data-centre development, while concerns about electricity costs, water supplies and community impact featured in several campaigns.
The Wesleyan Media Project also cautioned that its spending estimate was likely lower than the actual amount. Some advertisements carried no AI disclosure, while other uses of the technology could be difficult to identify through visual inspection alone.
