Researchers have tracked $80 million in AI-generated political spending during the 2026 midterms, uncovering a landscape of deepfakes and inconsistent disclosures.
The 2026 U.S. midterm campaigns have quietly crossed a threshold. moving from traditional talking points to a digital landscape where the line between reality and artifice has blurred. As campaigns pour money into television and social media. an $80 million surge in spending is now tied to a technology that can place candidates in rooms they never visited and situations they never navigated.
Working with data from AdImpact and student coders. researchers from the Wesleyan Media Project have identified 170 unique advertisements utilizing generative AI. These range from subtle visual enhancements to hyperrealistic deepfakes, turning the election cycle into a testing ground for synthetic media. In one instance, a Republican Senate candidate from Louisiana appears to drive a bus filled with undocumented immigrants. In another, a Republican candidate for governor in South Carolina walks arm-in-arm with drag queens. Other spots show figures like Liz Cheney. Mitt Romney. and Mike Pence carrying pitchforks on the White House lawn. or a fake Dr. Anthony Fauci running through a state fair wielding a giant syringe.
Republican sponsors—ranging from individual candidates to super PACs and 501(c) organizations—are driving this trend. They account for 80% of the identified AI ads and 83% of the total $80 million expenditure. While Democrats have historically prioritized regulatory frameworks. such as the 2022 push for the DISCLOSE Act. the current environment remains governed by a patchwork of state-level rules that often fail to keep pace with the technology.
The human element, however, is the primary target of these edits. Alexandria Ocasio-Cortez has appeared in at least five separate AI-generated spots. while other ads feature synthesized versions of Donald Trump. Nancy Pelosi. Barack Obama. and Kamala Harris. Even when AI is used merely to enhance visuals or generate background crowds rather than creating a full deepfake. the lack of clarity leaves the average viewer struggling to determine what is true. While 78% of registered voters favor banning AI content that makes deceptive claims, the current enforcement is inconsistent at best.
Transparency efforts remain erratic. Across 35 states, only 31% of the ads tracked included a disclosure, and the language used is far from uniform. One Georgia ad labeled itself as “manipulated or generated with artificial intelligence. ” while an Oklahoma spot claimed to be “political satire.” A North Carolina state Senate ad took a more casual approach: “You guessed it!. AI was definitely used to generate these silly video clips.”.
The gap between law and reality is stark. In states that have passed laws regulating AI in political advertising. only 29% of ads contain a disclaimer—a rate that actually trails the 32% found in states with no such regulations at all. Specificity remains a hurdle; for instance, Colorado’s law only covers candidate deepfakes, leaving other AI uses unregulated. In Minnesota. an ad featuring a synthetic version of Democratic Senate candidate Peggy Flanagan aired in May 2026. despite a state law banning political deepfakes. The question of whether it violated the statute—which requires the content to be “so realistic that a reasonable person would believe it depicts speech or conduct” that did not happen—remains unsettled.
As the cycle progresses. the challenge for voters is no longer just deciphering a candidate’s platform. but verifying the footage they see on their screens. For now. the regulation of AI in politics remains a map of disparate rules. voluntary disclosures. and the persistent. unanswered question of how much truth survives the edit.
AI political advertising 2026 midterms deepfakes Wesleyan Media Project political campaign spending