Wesleyan Media Project researchers have tracked nearly $187 million in AI-driven campaign ad spending, and most of those ads carry no AI disclosure
The 2026 midterms have a new star, and it isn’t a candidate. It’s synthetic video.
Campaign ads built from AI-generated clips and images, a genre now known as slopaganda, have become a defining feature of this election cycle. That leaves Instagram, YouTube and X with another moderation headache, and no federal rulebook to lean on.
The stakes are immediate. Election day falls on November 3, and the spending curve is still pointing up.
The numbers behind the slop
Researchers at the Wesleyan Media Project have been counting. By mid-October 2026, they had identified more than 220 AI-generated political ads running across 34 states, backed by over $137 million in spending.
A couple of weeks later, the count had grown. By late October, the tally reached approximately 289 ads and nearly $187 million in spending.
For context, the previous election cycle saw under $50,000 spent on this kind of advertising.
The spending is also lopsided. Republican candidates and aligned groups account for 80-88% of the ads and expenditures the researchers tracked.
Then there’s the transparency problem. Only about 27% of the tracked ads contain any form of AI disclosure, per the research findings, while roughly 70% carry no AI disclaimer at all.
Put plainly, most voters seeing these ads get no signal that what they’re watching was generated by software. That gap exists even though platform policies require such notices on election-related content.
Where the platforms stand
Each of the big three has some kind of policy on the books. Meta, which owns Instagram, uses automated labeling to flag synthetic content. Google, which owns YouTube, and X have both stated that they have rules in place for AI in political advertising.
The ads also don’t stay in one lane. Many of them run both on social platforms and on television. That splits oversight between tech companies with their own content policies and broadcast channels operating under different norms.
A word born in a research paper
The term slopaganda isn’t campaign-trail slang. It dates back to early 2025, when a group of AI experts used it in an academic paper.
The portmanteau mashes together two ideas: AI “slop,” the low-effort generated content clogging feeds, and propaganda paired with misinformation. The researchers used it to describe a new phenomenon blending the two.
A legal vacuum, filled state by state
There is no comprehensive federal law governing how AI can be used in political ads, or how it must be disclosed.
States have tried to fill the gap, with mixed coverage. Only 31 of the 50 states require some form of disclosure for political ads that include AI-generated content.
That means an ad that requires a disclaimer in one state can run without one across the border. For campaigns buying national digital inventory, the rules depend on the ZIP code of the viewer.
What this means
When roughly 70% of tracked ads carry no disclaimer anyway, critics have an easy data point to argue those policies aren’t working.
The spending trajectory matters too. Going from under $50,000 in one cycle to nearly $187 million in the next suggests campaigns see real value in the format.
The partisan skew adds a political wrinkle. With Republican candidates and aligned groups behind 80-88% of the tracked activity, any push for stricter federal disclosure rules could quickly turn into a partisan fight rather than a neutral tech-policy debate.
What to watch: the final Wesleyan Media Project tallies once the November 3 vote is in, and whether the 19 states without AI disclosure requirements decide to join the other 31.
