AI campaign satire puts disclosure under scrutiny before US midterms

US midterm campaigns are using conspicuously artificial videos to portray candidates ahead of the November 3 election, Al Jazeera reported on October 9. Examples from New York and Senate races across the country have put campaign satire and AI disclosure under scrutiny, with researchers questioning how fabricated performances could shape voters’ perceptions.

The reporting comes against a substantial spending baseline: the Wesleyan Media Project identified at least 164 AI-generated or AI-enhanced political advertisements in its September 4 assessment, carrying just under $80 million in estimated advertising spending. Those figures describe an earlier tracking exercise, not an updated October total.

New York campaign ads draw a disclosure dispute

According to Al Jazeera, Republican New York gubernatorial candidate Bruce Blakeman released AI ads depicting Governor Kathy Hochul and New York City Mayor Zohran Mamdani gardening together and collaborating to increase residents’ costs. Another synthetic portrayal showed Hochul discussing higher tolls and transit fares. Neither politician actually performed in those scenes.

Blakeman defended the material as clearly labelled AI satire, invoking an established tradition of parody in American politics. Al Jazeera reported that the advertisements carry an AI disclaimer for several seconds at the beginning. It also reported that New York Democrats filed a state elections board complaint alleging disclosure violations and misleading advertising.

The examples extend beyond New York. Al Jazeera described Ohio Democratic Senate candidate Sherrod Brown portrayed as a party guest who had overstayed his welcome, Texas Democratic Senate candidate James Talarico in a dress resembling a Sound of Music character’s, and Michigan Republican Senate candidate Mike Rogers as a superhero.

What Wesleyan’s AI advertising figures measure

Wesleyan’s September assessment provides a dated measure of identified AI advertising across broadcast television, connected television and digital venues. Its initial May assessment had located 33 advertisements with about $20 million in spending. The later tally represents growth in the number identified, but neither assessment is a complete census of campaign AI use.

Republican candidates and aligned groups accounted for 80 percent of the identified ads and 83 percent of their estimated spending. Democratic sponsors accounted for 11 percent of the ads and 14 percent of spending. The remaining advertisements included issue advertising and material supporting independent candidates.

Within that identified set, 69 percent of advertisements lacked a disclaimer disclosing AI use. Those ads represented 78 percent of the estimated spending. The percentages measure disclosure within Wesleyan’s sample; they do not establish that particular advertisements broke a law.

The project uses AdImpact spending estimates for each identified advertisement. Its release does not describe the money as the cost of producing AI content. Although the release’s headline says at least $80 million, its detailed assessment specifies just under $80 million, the figure used here.

Wesleyan’s September 4 release explains that identification combined AI disclaimers, news or AdImpact reports, and staff assessments confirmed by its team. The researchers warn that uses of AI can escape detection, making the tally an underestimate rather than a definitive account of prevalence.

Researchers debate labels and campaign judgement

Scott Babwah Brennen, director of New York University’s Center on Technology Policy, told Al Jazeera that visibly artificial campaign material can communicate “deeper truths rather than factual truths”. His concern was that such portrayals could influence perceptions of candidates’ character or ideology and affect voter trust.

Public Citizen’s Ilana Beller called for AI advertising disclosures to be as understandable as food labels, according to Al Jazeera. She also advocated greater awareness of those labels among younger voters and seniors as the campaign enters its final stretch and beyond.

Adam Rose, chief operating officer of Starling Lab, offered a different emphasis. He told Al Jazeera that some AI alterations amount to colour or voice correction and should not require government labelling. He opposed banning AI-generated advertisements and argued that campaigns’ decisions about using the technology reveal candidates’ judgement.

These are competing assessments of how campaigns should use AI and explain it to voters. Al Jazeera’s October 9 reporting does not measure changes in voting decisions caused by the advertisements. Wesleyan’s assessment tracks identified ads and spending, leaving their electoral effects unresolved as the November 3 vote approaches.