US political campaigns are increasingly using artificial intelligence in election advertising, from digitally altered scenes featuring well-known politicians to synthetic videos depicting opponents in situations that never occurred.
Research published by Wesleyan University's Wesleyan Media Project has identified 220 adverts created at least partly with AI tools during the 2026 election cycle, with estimated advertising expenditure exceeding $137 million (£102 million).
Researchers warn that the figures are likely to underestimate AI use because some applications of the technology are difficult to detect.
The analysis found that Republican candidates and groups sponsored 81% of the identified adverts, compared with 12% sponsored by Democrats. Republican-affiliated sponsors accounted for 88% of the estimated advertising expenditure. These figures relate to the adverts identified by researchers, rather than every political campaign's use of AI.
AI Political Ads Raise Questions Over Transparency and Campaign Tactics
One example comes from Oklahoma, where an advert produced by the Make Oklahoma Great Again political action committee (PAC) depicts Republican gubernatorial candidate Mike Mazzei embracing an AI-generated version of Hillary Clinton, while an artificial depiction of Donald Trump stands to the side.
The advert labels Mazzei a 'Clinton lover, Trump hater'. According to the New York Times, the group spent more than $200,000 (£149,000) purchasing television airtime for the commercial. The newspaper reported that the advert did not disclose the use of AI.
The Wesleyan Media Project found that just 27% of the identified adverts carried a disclaimer disclosing the use of AI, leaving approximately 73% without one. Erika Franklin Fowler, a professor of government and co-director of the project, told the New York Times: 'I think what we're seeing with AI has surprised even us.
Some of the ways in which AI is used, honestly, if there wasn't a disclaimer on the ad, it would be hard to know that AI was there.'
AI is not always used to fabricate an entire scene. Campaigns can also employ the technology to create backgrounds, generate crowds, alter existing footage or produce visual material that would otherwise require more time and money.
This can make it difficult to distinguish routine production assistance from synthetic content that could leave viewers with a false impression of a candidate.
Researchers have discussed whether differences in attitudes towards regulation and the perceived reputational cost of using AI could help explain the partisan gap. These remain possible explanations rather than conclusions established by the advertising data, which does not determine why the difference exists.
Voter Opposition and AI Election Laws Add Pressure on Campaigns
A survey of 1,568 likely 2026 voters commissioned by the American Association of Political Consultants Foundation found that 95% considered it unacceptable for a campaign to use AI to create a video showing an opponent saying something they had never said. Even when a disclosure was included, 93% continued to consider that practice unacceptable.
The results distinguish objections to deceptive political content from opposition to AI in every setting. Using software to help draft a script is not the same as creating a convincing video that attributes fabricated words or actions to a real person.
The survey does not establish how the identified adverts affected voting decisions, but it indicates that synthetic depictions of opponents may carry reputational risks for campaigns.
The regulatory picture is uneven. The National Conference of State Legislatures tracks state laws governing AI-generated material in election campaigns, including requirements to disclose synthetic content and restrictions on certain deceptive depictions. The rules differ by state, the type of media involved and the timing of publication.
Some laws have also faced legal challenges, making it important to check the current status and scope of the specific rules relevant to a campaign.
These differences create practical difficulties for advertisers distributing the same material across several states. A disclaimer that meets one state's requirements may not satisfy another state's rules, while some content may fall outside a law's scope altogether.
Campaigns and political groups therefore need to consider both the accuracy of their material and the disclosure obligations that apply where it is distributed.
For voters, checking original footage and consulting reliable reporting can help establish whether a scene is authentic. Disclosures can provide useful context, although their presence does not prove that footage is genuine, and their absence does not establish that it is fabricated.
As the November midterms approach, transparency and the distinction between genuine footage and fabricated scenes remain central questions for voters and regulators. The extent to which AI-assisted advertising influences electoral outcomes remains uncertain.
