AI political ads top $137 million: which state laws apply

Key takeaways

• The Wesleyan Media Project counted 220 AI-generated or AI-enhanced political ads, backed by more than $137 million, in its October 15, 2026 release. It found less than $50,000 of such spending in the whole 2024 cycle.
• Only 27% of those ads carried a disclaimer saying AI was used, according to the project.
• No federal law specifically regulates AI in political ads. The FEC says its existing fraud rule is "technology neutral" and decides cases one at a time.
• NCSL counted 31 states with election-deepfake laws in June 2026. Twenty-eight of them rely on disclosure, and Maryland, Minnesota and Texas ban certain deceptive deepfakes outright.
• Federal judges have blocked California's AB 2839 and AB 2655 and Hawaii's Act 191, which leaves about 29 state laws in force, according to Axios.

Are AI political ads legal in the 2026 midterms?

AI political ads have topped $137 million in the 2026 midterms, the Wesleyan Media Project found, yet no federal law bans them. The rules depend on the state: NCSL counted 31 states with election-deepfake laws in June, most requiring a label. Voting ends November 3.

In practice, a campaign can use AI to make an ad almost anywhere in the country. In most states that have acted, the obligation starts only when the ad is "materially deceptive," meaning it shows a real candidate saying or doing something that never happened, and it runs inside a set window before an election. Even then, most laws only require a clear disclosure. A few states go further and treat some deepfakes as a crime.

How much is being spent on AI political ads?

The Wesleyan Media Project is a nonpartisan political-ad research group at Wesleyan University in Connecticut. It has tracked AI ads in real time this cycle, and its count has grown with every update:

Source: Wesleyan Media Project releases of May 28, September 4 and October 15, 2026. Spending figures are AdImpact estimates used by the project.

According to the October 15 release, 81% of the AI ads were sponsored by Republican candidates or groups, and those sponsors accounted for 88% of AI ad spending. Democrats sponsored 12% of the ads, and independent candidates and groups sponsored the rest. The ads appeared in 34 states. "AI is making its way into the 2026 midterms in a big way this year," said Erika Franklin Fowler, the project's co-director.

The project says its figures are probably too low. It counts an ad only if the ad discloses AI use, if news outlets or AdImpact have reported it as AI-made, or if its coders flag it and the team confirms it. Ads it is unsure about are left out. Fowler told WSHU in October that the count is "absolutely an undercount," because little AI use is disclosed.

Examples reported by news outlets

The National News Desk (TNND) summarized reporting by The New York Times on October 16 and described several cases. One ad showing New York Mayor Zohran Mamdani and Governor Kathy Hochul promoting a fictional pill ran without an AI disclaimer. Senator Marsha Blackburn's campaign for Tennessee governor drew criticism for an AI video of Republican candidates dancing. Representative Thomas Massie of Kentucky, who lost his primary, said AI ads had falsely suggested he dated two Democratic colleagues. He wrote on X that "many older people voted against me" because of them. That is his own account, and Bullquill has not independently verified its effect on the result.

What the federal government does and does not regulate

No federal rule has been written specifically for AI in campaign ads. Two agencies have looked at the issue.

• Federal Election Commission (FEC). Public Citizen petitioned the FEC in July 2023, and the commission received more than 2,000 comments. On September 19, 2024, it declined to open a rulemaking. Instead it approved an interpretive rule saying the federal ban on fraudulently misrepresenting a candidate (52 U.S.C. 30124) applies "regardless of the technology used." The FEC calls that law "technology neutral" and says it will judge specific cases one by one. The interpretive rule is guidance and creates no new disclosure duty.
• Federal Communications Commission (FCC). In 2024, then-Chairwoman Jessica Rosenworcel proposed requiring on-air disclosure of AI content in broadcast political ads. The proposal would not have covered streaming or social media. Commissioner Brendan Carr voted against it and became FCC chairman in 2025. Bullquill could find no record that the proposal was finalized.

Online platforms have their own policies on labeling synthetic political content, but those are company rules, not law. They also vary between platforms.

Which state laws apply to AI election deepfakes

Different trackers count the states differently because they use different definitions. NCSL counted 31 states with election-deepfake laws in its June 2026 report, according to Axios. Public Citizen's tracker listed 30 on May 13, when Maryland signed its law, Biometric Update reported. Most of these laws were passed in 2024 or later. They fall into three broad groups:

Sources: NCSL via Axios (August 7, 2026); Biometric Update (May 15, 2026); Texas Legislature SB 751 bill analysis; Eighth Circuit opinion in Kohls v. Ellison; Bloomberg Law court reporting.

Which AI election laws have courts struck down?

California AB 2839 and AB 2655

Both laws were challenged in Kohls v. Bonta in the Eastern District of California. In August 2025, Senior US District Judge John Mendez struck down AB 2655, which would have required large platforms to remove reported deepfakes of candidates close to elections. He held that the law is preempted by Section 230 of the Communications Decency Act, Bloomberg Law reported. He then struck down AB 2839, which barred distributing deceptive media of candidates before elections, on First Amendment grounds. Mendez wrote that AB 2839 "discriminates based on content, viewpoint, and speaker," according to the Daily Journal.

Hawaii Act 191

On January 30, 2026, US District Judge Shanlyn Park permanently barred Hawaii from enforcing Act 191, its 2024 election-deepfake law, in a case brought by the satire site The Babylon Bee, Bloomberg Law reported. The judge found the law was based on content and speaker. She also found that its required disclaimer changed the speaker's message. The Hawaii attorney general's office said it was reviewing the decision.

Minnesota: still standing for now

Minnesota's ban has so far survived. On February 9, 2026, the Eighth Circuit upheld a lower court's refusal to temporarily block the law in Kohls v. Ellison, partly on standing grounds, and the case continues. X Corp. has filed a separate lawsuit arguing that the statute is too vague for platforms to follow. Neither challenge has stopped the law from applying this cycle.

What it means

For voters, a missing AI label does not mean an ad is real. By Wesleyan's count, most AI ads in this cycle carry no disclaimer, and in most states a disclaimer is only required for deceptive depictions of real candidates. Satire, parody and stylized AI imagery are often exempt or fall outside the definitions.

For campaigns and ad buyers, the rules are a patchwork. An ad that is legal with a label in one state can be a criminal risk in Maryland all year, or in Minnesota or Texas close to an election. The rulings in California and Hawaii show that courts are skeptical of broad bans and compelled labels. That makes more state laws likely targets for challenge after November 3.

At the federal level, the debate has shifted toward whether Congress should set a national AI framework that would preempt state laws. Public Citizen argues that this would wipe out state election safeguards. Supporters of preemption say the patchwork of state rules is a compliance burden. Neither outcome will arrive before this election, so for the 2026 midterms, state law and the FEC's case-by-case fraud rule are what apply.

For related election coverage, see Bullquill's reports on the fight over the Ramaswamy attack ad in Ohio, the US–Canada dispute over midterm interference claims, the Supreme Court's New Hampshire student ID voting case and the "No Kings" early-vote protests.

Sources

• Wesleyan Media Project: GOP Lead in Senate Spending Narrows; Dems Catch Up in Senate Broadcast Airings (Oct 15, 2026)
• Wesleyan Media Project: At Least $80 Million Spent on AI Ads This Cycle (Sep 4, 2026)
• Wesleyan Media Project: Past Two Weeks Sees $210 Million in Ad Spending; $20 Million Spent on AI Ads This Cycle (May 28, 2026)
• NCSL: Artificial Intelligence (AI) in Elections and Campaigns
• Axios: Voters face uneven AI deepfake protections (Aug 7, 2026)
• FEC: Commission approves interpretive rule on artificial intelligence in campaign ads (Sep 2024)
• AP via ABC News: FCC pursues new rules for AI in political ads (Jul 2024)
• Bloomberg Law: Musk-challenged California anti-deepfake law struck by judge
• Daily Journal: Judge strikes down California deepfake law as unconstitutional
• Bloomberg Law: Hawaii deepfake election law violates free speech, court finds
• US Court of Appeals for the Eighth Circuit: Kohls v. Ellison opinion (Feb 9, 2026)
• MediaPost: Musk's X Corp. sues Minnesota over deep fake law
• Texas Legislature: SB 751 (2019) bill analysis
• Biometric Update: Election deepfake laws spread across US ahead of 2026 midterms (May 15, 2026)
• WSHU via CT Mirror: A Connecticut report shows uptick in AI campaign ads nationally (Oct 8, 2026)
• The National News Desk: Millions of dollars spent on AI-generated political ads as midterms approach (Oct 16, 2026)