AboutSubscribe
Blog News

Campaigns spent $187 million on AI-generated ads in the 2026 midterms. Most carried no label

About 300 political ads made or enhanced with AI drew $187M in spending this cycle, and most had no AI disclaimer. What readers and publishers can check.

The Smattering Desk
· 5 min read
Illustration of a retro television and a smartphone with a floating mask between them, a ballot box and a checklist nearby

The first U.S. election in which generative AI was cheap and everywhere now has a hard number attached to it. Campaigns and outside groups spent more than $187 million running nearly 300 political ads that were generated or enhanced with AI tools, according to tracking by the Wesleyan Media Project reported by CNBC and, after the vote, by PolitiFact and Poynter. In the 2024 cycle the comparable figure was about $50,000.

The part that matters for anyone who reads news on a phone: most of those ads never said they were made with AI. In its September update, Wesleyan found that 69% of the AI ads it had identified, and 78% of the money behind them, carried no AI disclaimer.

That leaves the checking to editors and readers. Newsrooms that want provenance checks and AI-assisted triage built into their own CMS rarely train detection models themselves; they bring in an engineering partner, such as YuSMP Group's GenAI integration team, and keep a human editor on the final call.

How fast did AI ad spending grow this cycle?

Spending grew more than ninefold between Wesleyan's first and last counts:

  • May 2026: 33 ads, about $20 million.
  • September 4, 2026: at least 164 ads, just under $80 million.
  • Early November 2026: nearly 300 ads, more than $187 million.

Wesleyan calls the total "almost certainly an undercount." Its method relies on ads that disclose AI use, ads flagged by news reports or the ad-tracking firm AdImpact, and its own coding of broadcast, connected-TV and digital ads. There is no federal rule that requires a political ad to say it used AI, and for many ads AI use is hard to prove.

What did the AI ads actually show?

Some ads used AI to touch up real footage, for example by adding people to a crowd shot. Others invented scenes outright, PolitiFact reported. CNBC documented a fabricated video of a sitting governor released by an opponent with no disclaimer, and another ad whose AI disclosure appeared for about three seconds in the middle of the spot.

Whether any of this changed votes is unclear. Cayce Myers of Virginia Tech told PolitiFact that AI content "isn't necessarily persuading audiences," and voters were flooded with many other messages on other platforms. Christina Walker of the University of Georgia described AI ads as "the new normal."

Who is supposed to label AI ads?

Labeling falls to a mix of states and platforms, and the rules differ by channel.

  • States: 31 states have laws that regulate deepfakes in political messaging, according to PolitiFact.
  • YouTube and Google: prominent labels on realistic AI content, election-advertiser verification and "paid for by" disclosures.
  • Meta: "paid for by" labels, a public ad library and, since June, an "about this ad" note on content its systems detect as AI-made.
  • X: advertisers must follow applicable law; the platform adds little of its own.

Satire sits in an odd spot here. Courts have long protected obvious parody, as we covered in our piece on Hustler v. Falwell, and satire only works when the reader can tell it is a joke. A realistic AI ad with no label is built to do the opposite.

What can readers and newsletter writers do?

Treat a striking political clip the way you would treat a screenshot: as something cut off from its source until proven otherwise.

  • Find the sponsor. Look for the "paid for by" line, then search for the ad in the Meta Ad Library or the Google Ads Transparency Center.
  • Look for the original footage. Search for the scene outside the ad. If it appears nowhere else, treat it with caution.
  • Check the disclosure placement. A label that shows up for a few seconds mid-spot is easy to miss on a muted phone feed.
  • If you embed clips in a newsletter, link to the source and say plainly when footage is AI-generated or AI-enhanced.

The same habits behind our sixty-second check for telling satire from fake news apply here: slow down before you share.

Sources for this report

From the blog of The Smattering, where we write plainly about satire, media literacy and how we work. More posts.

Get the next one first

Free satire in your inbox once or twice a week.

More from the blog

All posts
Illustration of a magnifying glass over a newspaper next to a phone
Media literacy

How to tell satire from fake news in under a minute

Satire wants you to get the joke. Disinformation needs you to miss it. Here is how to check which one you are reading.

6 min read
Illustration of a smartphone with envelopes flowing into it next to a newspaper and a notification bell
Tech

Does an independent newsletter need its own app? A practical guide

Most small newsletters do not need an app. Here is when one pays off, what it really costs, and the cheaper options to try first.

9 min read
Illustration of a columned courthouse
Law

Satire and free speech: what Hustler v. Falwell still means

In 1988 the Supreme Court decided, unanimously, that public figures cannot win damages just because a parody hurt their feelings.

6 min read