As the midterm elections approach, campaigns and PACs have already spent over $137 million on AI-generated political advertisements, including a controversial 85-second video featuring Texas Senate candidate James Talarico reading his old tweets aloud and delivering commentary he never actually made. The Texas Senate Republicans’ campaign arm produced the video, which a deepfake expert described as one most viewers would not recognize as artificial, despite a faint “AI GENERATED” label in a corner.
The Rise of AI-Generated Political Ads
Between now and November 3, any clip of a candidate on television or in your feed may be fabricated, and most carry no AI label whatsoever. The Wesleyan Media Project reported the $137 million figure, while the nonprofit TrueMedia.org, founded in 2024, developed a detector to analyze over 60,000 election-related media files. However, the tool’s early deployment meant it could not fully assess the scale of AI ad usage during the 2024 cycle, when spending on such ads was under $50,000. In January 2025, the founder said, “It did not materialize in 2024. That doesn’t mean that it’s not coming.”
In 2026, the trend has shifted dramatically. American campaigns and PACs are increasingly using AI-generated content, both on broadcast television and digital platforms. Many ads fall into harmless categories, such as animated still photos or synthetic families replacing costly stock footage. Veteran campaign ad creator Fred Davis noted that what once cost $2 million to film now costs roughly $200,000 to generate, with audiences remaining undeceived.
Deceptive Content and Voter Vulnerability
Other AI ads push boundaries. A Make Oklahoma Great Again PAC ad paired gubernatorial candidate Mike Mazzei with Hillary Clinton under the claim “Clinton Lover. Trump Hater,” without an AI label, costing over $200,000 for Oklahoma TV airtime. Similarly, a Minnesota ad depicted Senate candidate Peggy Flanagan reaching for cash. These ads aim to be believed, not dismissed as satire. A 2021 study found that 47% of participants believed a fabricated scandal in a deepfake of Elizabeth Warren, while 43% accepted a fake headline.
Read Also: Microsoft’s Comms Chief Steps Down After 30 Years
Voters struggle to detect fakes. A Utah Valley University study found only 16% correctly identified deepfakes, with media literacy offering little protection. Current laws, including statutes in 31 states, have limited impact. Labels appear in 29% of AI ads in regulated states and 32% in states without laws, yet 37% of viewers forget seeing required disclaimers, and only 21% connect “manipulated by technical means” to AI. Labels also reduce trust in the candidate, prompting campaigns to omit them.
Regulatory Challenges and Proposed Solutions
The proposed framework focuses on the person depicted, not the technology. A real person must consent to their representation; otherwise, ads require on-screen and audio labels or cannot air. This straightforward test could be applied by station managers in minutes, addressing debates over defining “deceptive” content. Federal legislation like H.R. 9922 seeks standardized disclaimers, replacing the patchwork of state laws.
Public support is growing. A Scripps News poll found 43% of registered voters had already encountered a deceptive AI ad, with 77% of Democrats and 79% of Republicans favoring regulations. As the 2028 presidential election looms, experts warn of escalating deception risks if rules are not established by then.
Implementation and Industry Response
Georgetown University researcher Sejin Paik highlighted the challenge of deceptive content on television. Paik noted that viewers watching ads on TV lack the opportunity to react or scrutinize context, making clear labeling essential. The consultant trade group has endorsed federal standards to replace the patchwork of state laws, arguing that uniform rules would streamline compliance. H.R. 9922, pending federal legislation, seeks to codify these disclaimer requirements, potentially shaping how political media is regulated nationwide.