Midterms: $137M on AI-Made Ads, $78.5M From AI Groups

Campaigns and groups have spent more than $137 million on 220 political ads made at least in part with AI tools in the 2026 US midterm cycle, the Wesleyan Media Project reported on 15 October 2026. For the whole 2024 cycle, it found less than $50,000.

A separate CNBC analysis of Federal Election Commission filings, published on 27 October 2026, counts about $78.5 million spent on federal candidates this cycle by political groups on both sides of the AI industry’s own policy fight. The two figures measure different things, ads made with AI and money from AI-aligned groups, so we do not add them.

Video explainer · AI in the midterms

A 16:9 video explainer narrated from the published headlines of this piece and four related FourWeekMBA pieces.

AI-Made Ads: $137 Million

The Wesleyan Media Project says it now counts 220 ads generated at least in part with AI tools, on which more than $137 million has been spent, and that this is “almost certainly an undercount given that it is difficult to detect uses of AI.”

“AI is making its way into the 2026 midterms in a big way this year. For comparison, we identified less than $50,000 in spending on AI-generated ads in the 2024 cycle,” said Erika Franklin Fowler, co-director of the project. “Much of the AI use is not disclosed through any type of disclaimer.”

The project’s count has grown release by release. On 28 May 2026 it identified at least $19.7 million spent on 33 AI ads; its 4 September release was titled “At Least $80 Million Spent on AI Ads This Cycle”, with a minimum of 164 ads; on 15 October it reported more than $137 million on 220 ads.

By sponsor, the project says 81 percent of the AI-generated ads were sponsored by Republican candidates or groups and 12 percent by Democrats, with the rest from independent candidates or groups on issues that are not ostensibly partisan. Republican-affiliated sponsors account for 88 percent of the spending on these ads. The ads appeared in 34 states, and 27 percent of them contain a disclaimer.

The spending figures are AdImpact estimates. The project says it relies on its coders, the presence of disclaimers and AdImpact’s ad tagging, and leaves out ads when it is unsure whether AI was used, so its counts are likely on the low side.

The AI Industry’s Own Money: $78.5 Million

CNBC identified federal committees aligned with two camps and totalled their FEC-reported independent expenditures for or against congressional candidates. Groups favouring faster AI development and lighter, more uniform rules “have spent roughly $8 million in general-election congressional races, besting groups pushing for stronger safeguards and more room for states to regulate that have spent about $6.9 million,” CNBC wrote.

In the primaries, CNBC says, Leading the Future’s network reported roughly $26 million and the safeguards-focused groups about $32.1 million, with another roughly $5.5 million in runoff and special-election spending. About 74 percent of the roughly $78.5 million went to primary races, according to CNBC’s analysis. The analysis excludes state races and nonprofit advocacy spending.

Leading the Future launched in August 2025 with more than $100 million in commitments from backers including Andreessen Horowitz and OpenAI President Greg Brockman and his wife, Anna, CNBC reported. On the other side, Anthropic committed $20 million to Public First Action in February and another $20 million in July.

The two camps met most expensively in New York’s 12th District Democratic primary. CNBC says Think Big spent $8.3 million trying to defeat state Assemblymember Alex Bores, who authored state AI safety legislation, and the Anthropic-backed Jobs and Democracy PAC countered with about $15 million to support him. Together the two spent $23.3 million, about 40 percent of the camps’ primary spending; Bores lost to Micah Lasher, 39 percent to 35 percent.

Some backers have pulled back. CNBC reports that the Brockmans gave $25 million to Leading the Future and later abandoned plans for another $25 million, per The New York Times, and that groups Meta funded returned $25 million this summer, per IRS and state filings. “Something’s clearly not working,” OpenAI CEO Sam Altman told Politico. “I don’t think spending money on politics is what’s going to make people feel better about it.”

CNBC’s own All-America Economic Survey found 52 percent of registered voters said companies should pause development of advanced AI until Congress passes regulations, compared with 32 percent who favoured moving ahead.

What the AI Labs and Platforms Say

OpenAI’s 20 October update says it now applies image stylization to requests for images depicting election interference or that may discourage voting, producing “cartoonified” illustrations rather than photorealistic images. CNET reported on 26 October that every request it made for an image of Immigration and Customs Enforcement officers outside a polling location came back in a cartoon style.

“The problem with deepfakes isn’t social commentary, it’s deception. So our aim is to degrade the evidentiary value of an image, while preserving room for illustrations and social commentary,” Josh Lawson, OpenAI’s head of elections, told CNET.

The Guardian reported on 27 October that experts say social media feeds are awash with AI-driven disinformation ahead of the vote. “You can run an information operation now with quality that was impossible in 2016 for super, super cheap,” Frances Haugen, the former Facebook data engineer turned whistleblower, told the Guardian. Meta and X did not respond to the Guardian’s requests for comment; YouTube said all its users are subject to the same guidelines.

Connecting the Coverage

On 23 October we reported AI in the midterms: 80% fake-image rate and $2.5M in PAC ads, from a NewsGuard audit in which ChatGPT Images 2.5 complied with 16 of 20 prompts for compromising images of candidates.

On 24 October we reported that Pennsylvania signs AI ad label law with fines up to $3,000, and that Grok and OpenAI’s GPT split on midterm races: 92% vs 97%, in a study reported by TechCrunch.

On 25 October we reported that Claude pulled paid Georgia vote data without being asked, per Axios, and on 26 October that a campaign pitch prices a ChatGPT ‘category win’ at $200K, per Reuters. What’s new here: the two money counts side by side, $137 million on AI-made ads and $78.5 million from AI-aligned groups, a week before the vote.

The Structural Read

As we read it, the two counts point in different directions on timing: Wesleyan’s tracked AI-ad spending rose from at least $19.7 million in May to more than $137 million in October, while CNBC says about 74 percent of the AI groups’ federal money went to primaries.

As we read it, the AI industry’s money is split against itself: CNBC counts roughly $8 million from groups favouring faster AI development and about $6.9 million from safeguards-focused groups in general-election races.

As we read it, the platforms’ answers are narrow by design: OpenAI says it now turns images depicting election interference or that may discourage voting into cartoon-style illustrations, while the Guardian reports experts saying AI-driven disinformation is slipping through.

Three Implications

FOR READERS OF POLITICAL ADS The Wesleyan Media Project says 27 percent of the AI-generated ads it tracked contain a disclaimer.

FOR THE AI INDUSTRY’S POLITICAL SPENDING CNBC reports the Brockmans abandoned plans for another $25 million to Leading the Future, per The New York Times, and groups Meta funded returned $25 million this summer.

FOR AI IMAGE TOOLS OpenAI says requests for images depicting election interference or that may discourage voting now produce stylized, cartoonified illustrations rather than photorealistic images.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Supercycle.

The framework places governance at the top of a nine-layer AI stack and argues that “The supercycle has a ceiling now, and the ceiling is political.”

As we read it, the midterm spending, both the AI-made ads and the AI groups’ independent expenditures, is one place where that governance layer meets voters. The framework was not written about these elections.

What Is Not Established

Wesleyan’s $137 million is its own floor and, by its own account, an undercount. CNBC’s $78.5 million covers federal candidates only and assigns committees to camps by its own analysis. The two counts overlap in theme, not in money, and neither shows what effect the spending has on any race.

The Bottom Line

One week before the 3 November vote, AI shows up in the midterms twice in the money: more than $137 million spent airing ads made with AI tools, per the Wesleyan Media Project, and about $78.5 million spent on federal candidates by groups on both sides of the AI industry’s policy fight, per CNBC’s FEC analysis.