AI-generated political ads cross $137 million in the 2026 midterms

The numbers

Firstpost reported on 27 October 2026 that more than $137 million has been spent on political ads made at least partly with AI tools in the 2026 midterm cycle, according to the Wesleyan Media Project, a research group based at Wesleyan University that works with the ad-tracking firm AdImpact. Its latest tally, covering ads identified through October 11, found 220 AI-generated or AI-enhanced political ads in 34 states.The scale shift is dramatic: in the 2024 cycle, the same researchers found less than $50,000 in such spending.

Who is spending

According to Wesleyan's data, Republican candidates and groups sponsored about 81 percent of the identified AI ads and accounted for roughly 88 percent of the spending. The partisan split is notable but should be read against how the media projects classify ads: some outlets also found substantial undisclosed AI use across the political spectrum.

CNN separately found at least $38 million spent on ads that appear to use AI imagery without disclosing it. The Times reported some campaigns use AI mainly to create stock-style footage and cut production costs rather than to deceive, which the Wesleyan database reflects: nearly half of the AI ads identified include fabricated depictions of real politicians.

The disclosure problem

Only a minority of the AI ads carry an AI disclaimer: research cited in the report puts the figure at about 27 percent. That means most AI-generated or enhanced political advertising is being shown without a label telling viewers the content was machine-produced.

The regulatory gap is the story's core: the FEC has rules on disclosure of who paid for ads, but the AI-advertising disclosure regime is partial and inconsistent across platforms and states. Pennsylvania recently passed a law requiring advertisers to label AI-generated images, video and audio in ads, but most states have no such requirement.

Fabricated depictions

Nearly half of the identified AI ads include fabricated depictions of real politicians, per the Wesleyan database. A concrete example: on September 8, Ken Paxton's campaign released a 30-second AI-generated attack ad mixing synthetic speech with an authentic 2021 James Talarico comment — an ad that did include an AI disclaimer.

The range of use runs from cheap production tools to deceptive fabrication, which makes a single regulatory answer difficult. The same technology that helps a small campaign produce broadcast-quality footage affordably can also be used to create fake video of an opponent saying something never said.

What it means

The $137 million figure signals that AI has moved from the edges of campaign advertising to its center in a single cycle, faster than the regulatory system has adapted. Voters in the closing weeks of the midterms are now exposed to machine-generated and machine-enhanced political content at unprecedented scale, much of it undisclosed.

For platforms, the numbers raise the stakes on content labeling and provenance tools; for candidates, they create an arms-race dynamic where declining to use AI can mean ceding production cost advantages; and for voters, the practical advice is unchanged but more urgent — check the source, look for disclaimers and treat dramatic late-campaign videos with skepticism.

What to watch

Watch whether the FEC or Congress acts on AI ad disclosure before or after the midterms, and whether platforms tighten labels in the final weeks.

Watch the final Wesleyan tally after Election Day, since the October 11 cutoff predates the closing spending surge.

Watch whether undisclosed AI ad spending grows beyond the CNN-identified $38 million as campaigns finalize their media buys.

The regulatory gap

The FEC requires disclosure of who paid for political ads but does not currently mandate labels for AI-generated content across all media. The result is a mosaic: some platforms label AI content voluntarily, some states like Pennsylvania require it, and federal candidates are governed by different rules than independent expenditure groups.

The report's 27 percent disclaimer figure captures the gap in practice: most AI-generated or AI-enhanced political ads are running without any label telling viewers the content was produced with machine assistance. There is no single standard for what counts as 'AI-generated' either — an ad with AI-synthesized voice, an ad with AI-generated imagery and an ad with an AI-assisted edit are all being tallied together.

The production economics

The spending surge is as much about cost as about deception. AI tools let campaigns generate stock-style footage, localize ads, create variations and cut production costs to a fraction of traditional video production. The Times reporting that some campaigns use AI mainly to save money is consistent with the Wesleyan data showing a range of use.

That is why a blanket ban on AI ads is unlikely: much of the $137 million is legitimate production efficiency, and campaigns across the spectrum benefit. The hard regulatory problem is the subset — nearly half of identified ads with fabricated depictions of real politicians — where the technology is used to deceive rather than to economize.

Caveats

The Wesleyan Media Project figures are a research estimate based on identification methods that can miss AI-enhanced ads that leave no obvious trace; the actual total is likely higher. The partisan breakdown (81 percent of ads, 88 percent of spending Republican) reflects identified ads and should not be read as a complete measure of all AI ad use.

The disclosure figure of about 27 percent comes from research cited in the report; definitions of 'AI-generated' vary across studies, so comparisons between cycles and between studies are approximate.