Wesleyan team tracked $80 million in AI political ads

The 2026 U.S. midterm campaigns mark the first cycle in which AI-generated political ads are regularly appearing on televisions and social media feeds, according to researchers who have studied political advertising through the Wesleyan Media Project since 2010. Working from media reports, student coders and AdImpact, a firm that tracks political ad spending, they catalogued almost 170 unique AI ads across the cycle.

The ads themselves range from hyperrealistic deepfakes — synthetic video showing people doing things they did not do — to subtle visual enhancements that do not necessarily look different from pre-AI advertising. The researchers note that several ads featured deepfakes of figures including Donald Trump, Nancy Pelosi, Barack Obama, Kamala Harris and Alexandria Ocasio-Cortez, and that AI was also used to generate crowds or constituents. A separate strand of the data concerns disclosure: some advertisers volunteer AI labels even where no law requires them, while others omit disclaimers even where state law calls for them.

The researchers offer only speculation about why Republicans dominate AI use in this cycle, pointing to broader party differences over campaign regulation — Democrats have generally favored limits on campaign spending and required disclosures, while Republicans have tended to favor a free-market approach. What the outcome hinges on, it seems, is not the presence of state laws alone: coverage varies by state, some laws target only candidate deepfakes, and at least one law took effect after an ad had already aired. The test is likely to be whether transparency rules, enforceability, and the design of disclosures can be made to matter to voters in practice.