AI political ads have moved from curiosity to a line item in the 2026 midterms. Campaigns and outside groups have spent more than $137 million on ads made or enhanced with artificial intelligence, and most of those ads do not carry a clear AI label. Election Day is Tuesday, November 3, and broadcast, cable and streaming spots can keep running until the polls close.
The New York Times reported the spending and labeling figures, which trace back to the Wesleyan Media Project, a research group that tracks political advertising. Other outlets carried the same numbers, including AI Weekly. AFP re-reported the Wesleyan data on October 26. The data is about a week old, but the disclosure gap it describes is live this weekend.
How fast AI political ads grew in the 2026 midterms
The count has risen quickly. A September 4 Wesleyan report tallied 164 ads and just under $80 million. Six weeks later, the October 15 update counted 220 ads in 34 states and more than $137 million, according to Breezy Scroll.
| Report date | AI-made or AI-enhanced ads | Reported spending |
|---|---|---|
| September 4 | 164 | Just under $80 million |
| October 15 | 220 in 34 states | More than $137 million |
The dollar figure needs context. Breezy Scroll notes that the total counts airtime bought for ads that used AI somewhere, so $137 million is not what it cost to make the ads. A narrower count from AdImpact, limited to ads with an identifiable disclosure, comes in at just over $13 million.
Production costs may be falling. A Hollywood ad executive quoted by the New York Times described a commercial’s cost dropping from $2 million to $200,000. That gap may help explain why smaller campaigns and outside groups can now produce more spots.
The Wesleyan team counts an ad as AI-made if the ad discloses it, if news outlets or AdImpact reported it, or if its coders flagged it and the team confirmed it. That method means the count is a floor. Researchers believe the true number of AI-using ads exceeds what any current dataset captures.
Most of the ads carry no AI label
About 70% of the ads carry no AI label, according to the New York Times reporting. Gizmodo, citing Wesleyan, found that only 27% included a disclaimer. IBTimes reported 73% with no disclaimer. Yahoo’s coverage put the share that disclosed AI use at 31% in a narrower set.
Those numbers measure different samples and definitions, so they should not be merged into one statistic. What they share is a direction: most ads are not telling viewers that AI was involved.
Erika Franklin Fowler, co-director of the Wesleyan Media Project, has said that some AI applications are hard for viewers to detect without a disclosure label. That is the core of the problem. A viewer cannot judge a clip they cannot identify as synthetic.
The most troubling category may be deepfakes. Nearly half of the AI-generated ads the researchers identified feature deepfake depictions of real politicians, according to Gizmodo’s coverage. IBTimes described fake “Death to America” and “Celebrate 9/11” scenes that target candidates.
Which parties and races are involved
Republican-aligned groups lead the spending. More than 80 percent of the reviewed ads came from Republican campaigns or Republican-aligned PACs, per the Wesleyan Media Project’s broader dataset, as reported by Gadget Review.
Texas offers one of the most specific examples. An analysis by Etzioni, reported by IBTimes, found that a National Republican Senatorial Committee video showed Democratic Senate candidate James Talarico reading old tweets and voicing comments he never made. Talarico’s own campaign has drawn a contrast in its ads. One, as Gizmodo described it, ends with the words “Made by humans.” That is an ad line, not a policy statement.
New York adds a legal test. The New York State Democratic Party filed a complaint with the state Board of Elections over a set of advertisements tied to Blakeman. The complaint alleges a violation of a state law requiring materially deceptive political communications to carry an AI label. No enforcement outcome had been reported in the sources reviewed.
Who is paying for the AI-industry ads
The money behind these ads ties back to the AI industry. Leading the Future had $39 million banked at the end of last year and is wading into races through a pair of connected groups, according to NBC News. The group has received at least $20 million from the AI company Anthropic and has called for more significant AI regulation.
NBC News’s analysis is that ads funded by AI groups are flooding the 2026 election while mostly focusing on issues other than AI. Separately, a Wired report found that Think Big paid $20 million to a Nevada company that appears to operate from a post office box.
What the disclosure rules require right now
No federal requirement exists to label AI in political ads, and the reporting reviewed contained no pending federal bill. That leaves most of the rules to the states. About 30 states require AI disclosures for political ads, according to Gizmodo, and the requirements differ.
The Wesleyan Media Project proposed on-air AI disclosures for TV and radio political ads in July 2024. No such requirement covers this year’s ads.
Courts are also testing how far state AI laws can go. An appeals court paused Minnesota’s AI nudification ban for xAI, a reminder that state rules often face legal challenges. USA 24 covered that ruling.
How to check a political ad before you vote
- Look for an AI label. A missing label does not prove an ad is authentic, given the gaps described above.
- If a candidate appears to say something unfamiliar, check the candidate’s own website and social channels before sharing.
- Check local news for coverage of the ad, especially if it attacks a candidate.
- Check your state election agency’s rules on AI-disclosed political ads. This article did not verify individual state rules, so confirm them directly.
- Remember that ads can run until polls close on November 3, so a clip you see on Monday may still air on Tuesday.
What comes next
After Election Day, watch three things: whether the count of AI ads grows as researchers gather more data, the outcome of the New York complaint, and whether any federal lawmaker moves on labeling. The disclosure numbers will also vary depending on who does the counting, so compare sources carefully.
For the chatbot side of election information, USA 24 has reported how Copilot and ChatGPT cited a flawed site in midterm voter guides. That problem and the ad problem share a root. Voters often cannot see how AI shaped what they read or watched.
