How much AI political ads have cost this midterm cycle
Campaigns and PACs have spent more than $137 million on AI-generated or AI-enhanced ads this midterm cycle, and most of those AI political ads carry no label telling viewers that artificial intelligence was involved. The figures come from new data from the Wesleyan Media Project and AdImpact, reported by The New York Times on October 16 and 17, a few weeks before the November midterms.
The count is broad. It includes ads that are AI-generated or AI-enhanced, including ones that are not fabricated. Campaigns have used the tools for ordinary production work: animating still photographs, generating crowd scenes, recreating childhood settings and replacing conventional stock footage. A campaign that swaps a stock clip of a farm for a generated one uses the same technology as a campaign that puts an invented opponent on screen, but the risk to voters is not the same.
Money is also concentrating in a few places. Think Big, part of the Leading the Future network, a pro-AI super PAC network, reportedly paid about $20 million to Lantern Production Consultants. Wired found that Nevada company listed only at a post office box, according to AI Weekly.
Gizmodo describes the 2026 total as a massive jump from the 2024 cycle. The reporting does not include a 2024 dollar figure, so the size of that jump cannot be measured from these sources.
Most AI political ads never say so on screen
The disclosure gap is the most concrete angle for voters. According to one summary of the Wesleyan data, about 70 percent of ads in the broader review carried no AI label. In the narrower set, only 31 percent disclosed AI use.
Those two numbers describe different groups of ads, so they should not be mixed or compared as if they measure the same thing. Gizmodo's headline puts the share with a disclaimer at 27 percent, which uses yet another denominator. Readers who see all three figures are looking at three measurements, not one trend.
What the numbers share is the practical result. Most AI-made political ads in this cycle do not say so on screen, so a viewer cannot rely on a label to judge what they are watching. The viewer has to decide from the footage alone, which is a hard position to be in when a clip is built to provoke a reaction.
Wesleyan researchers say their counts likely understate the total, because ads in states without disclosure rules are harder to detect. If that is right, the places with the weakest rules may also hold the least visible ads.
The New York race at the center of the story
The most-cited example comes from New York. A Bruce Blakeman campaign video fabricated New York Mayor Zohran Mamdani and Gov. Kathy Hochul appearing together in a synthetic advertisement for a fictional "political delusion pill." The ad used their voices and likenesses without any AI disclaimer, according to the Times.
Blakeman is a heavy user of the format. The Times reports at least 22 AI-generated attack ads on his social feed since April, with only a few carrying a disclosure.
The legal fight is already underway. In August, the New York State Democratic Party filed a complaint with the state's Board of Elections over Blakeman's use of AI in advertising. The party claims he violated a state law requiring political communications that are "materially deceptive" to be labeled as AI-generated. The sources do not report the complaint's status, so whether the board has acted is unknown.
Other examples in coverage follow the same pattern. One Oklahoma ad shows a synthetic Democratic candidate embracing Hillary Clinton. Fabricated versions of Michigan Senate candidate Abdul El-Sayed and James Fishback also circulated in the reporting.
The Times quotes Bruce Schneier, a security technologist who teaches at Harvard and the University of Toronto, saying, "The problem isn't A.I." Read next to the New York complaint, that line points at how the tools are used and whether voters can tell, not at the technology alone.
Which states require a label, and what the rules look like
Roughly half of U.S. states require campaigns to disclose AI-generated or manipulated material in political advertising, according to CNN. The rules differ widely in scope, so the same ad can comply in one state and fall short in another.
Michigan's rule is unusually specific. It requires a three-second voiceover acknowledging AI use, alongside a four-second written disclosure. That gives a campaign a concrete standard to check against a finished spot, which is more useful than a general instruction to label AI content.
National campaigns buy ads across state lines, so a single spot may need to satisfy several sets of rules at once. The New York case shows the route available so far: a complaint to a state board under a state statute. The disclosure gap raises compliance questions under laws like New York's, and the answer depends on whether a given ad counts as "materially deceptive."
What the data can and cannot tell you
The numbers in this story come from different denominators, and outlets describe them in different units. One report describes the research as covering roughly 220 AI-generated ads. The sources do not explain how that ad count relates to the dollar total, so readers should not treat the two figures as interchangeable.
The 2024 comparison is also missing its anchor. Without a 2024 dollar figure, the claim of a massive jump is a directional judgment rather than a measured change.
The firmest figures are the spending total and the share of unlabeled ads. Both come from research whose authors say the problem may be larger than their counts show.
How to check an AI political ad before you share it
Start with the source. Treat emotionally charged candidate videos as unverified until you find the same footage on the candidate's official channel or in a reputable news outlet. A missing label does not prove a clip is authentic, and a label only tells you the creator disclosed a tool, not that the content is accurate.
Next, check your state's rules. Roughly half of states require AI disclosure in political ads, so look up your state election office's guidance on what a compliant ad must say and how it must appear. Michigan's voiceover-and-text standard shows how specific those requirements can get.
If a clip looks deceptive, report it. Send it to your state election office, and use the platform's reporting tool if you found it on social media. Include the link, the date you saw it and the account that posted it.
These are general steps rather than conclusions from the reporting, but they respond directly to the gap the data describes: viewers are being asked to judge footage the creator did not label.
What comes next for federal and state rules
Federal action has been limited so far. Trump has rejected calls for more AI regulation, arguing it would hold the United States back against China, according to Bloomberg as reported by The Next Web. That leaves states to set most of the rules, which produces the patchwork described above.
The American Association of Political Consultants, a trade group, is pushing for federal rules that would override the state patchwork. Julie Sweet, the group's director of advocacy and industry relations, said some operatives argue that labels can backfire.
States are not all standing still. Massachusetts is weighing independent safety tests for frontier AI models, a different kind of oversight that shows lawmakers moving beyond disclosure. In Washington, Cantwell and Rounds have urged the White House to build a coordinated AI response.
Three things bear watching. The first is whether the New York Board of Elections acts on the August complaint. The second is whether Congress sets a federal disclosure standard or preempts state laws, a question the reporting leaves unresolved. The third is whether Wesleyan and AdImpact publish updated counts that settle the denominator problem before Election Day. The next few weeks will show whether campaigns label more of what they produce or whether the gap stays as wide as the first data suggests.
