The 2026 midterms just became the first election where synthetic reality costs eight figures — and Republicans are buying 3x more of it than Democrats.
The Summary
- Researchers tracked $80 million spent across 170 unique AI-generated political ads in the 2026 midterms, the first U.S. election cycle with widespread AI ad deployment
- Republican campaigns and PACs are using AI-generated content at roughly 3x the rate of Democratic sponsors
- Most ads aren't disclosing AI use due to fragmented state-level legislation
- Content ranges from hyperrealistic deepfakes of politicians (Trump, Pelosi, Obama, Harris, AOC) to subtle visual enhancements and AI-generated crowd scenes
The Signal
The Wesleyan Media Project has been tracking political advertising since 2010. They've never seen anything like 2026. Nearly $80 million has flowed into AI-generated political ads this cycle, across 170 distinct creative executions. That's not a test budget. That's a category.
The partisan split is the first real tell. Republican sponsors are deploying AI-generated content at roughly three times the rate of Democratic campaigns. Not because Republicans are more tech-forward — they're not — but because the content strategy diverges at the foundation. Conservative ads are leaning into synthetic scenarios that would be expensive or impossible to film: deepfakes of AOC appearing in five separate ads, fake crowds at rallies, hypothetical scenes like Liz Cheney and Mitt Romney carrying pitchforks on the White House lawn.
"AI use spans from hyperrealistic deepfakes to subtle enhancements."
The deepfake deployment is surgical and repetitive. Alexandria Ocasio-Cortez shows up in at least five ads as a synthetic character. A Louisiana Senate candidate is shown driving a school bus full of undocumented immigrants. A South Carolina gubernatorial candidate walks arm-in-arm with drag queens. A fake Anthony Fauci runs through a state fair wielding an oversized syringe. These aren't accidents. They're narrative shortcuts — creating visceral, emotional moments that bypass the cost and logistics of traditional production.
But the hyperrealistic stuff is only part of the story. Most of the $80 million isn't going to obvious deepfakes. It's going to visual enhancements, crowd generation, and scene augmentation that sits just below the threshold of detection. AI-generated constituents. Synthetic rally attendees. Enhanced backdrops. This is where the real money lives — in making cheap content look expensive, and sparse support look massive.
Key deployment patterns:
- Deepfakes concentrated on high-recognition political figures (Trump, Pelosi, Obama, Harris, AOC)
- AI-generated crowds used to manufacture scale and enthusiasm
- Visual enhancements to polish low-budget productions into broadcast-quality spots
The disclosure problem is structural, not intentional. There's no federal requirement to label AI content in political ads. State-level legislation exists in patchwork form — some states require disclosure, most don't, enforcement is inconsistent. The result: most voters are seeing synthetic content with no marker that it's synthetic. Not because campaigns are hiding it (though some are), but because there's no uniform rule forcing transparency.
This is the first election where generative AI tools are cheap enough and good enough to be a standard line item in media budgets. The $80 million figure isn't the ceiling. It's the floor for 2028.
The Implication
Political advertising just crossed into a new production paradigm. When synthetic content costs a fraction of traditional video production and delivers comparable emotional impact, every campaign will build AI generation into their media plans. The 2026 spending level — $80 million — will look quaint in two years.
The disclosure gap is the bigger long-term risk. Voters can't evaluate synthetic content if they don't know it's synthetic. Federal legislation will eventually catch up, but not before several more election cycles run under fragmented state rules. Until then, campaigns will optimize for what works, not what's labeled. Watch for synthetic content to migrate from shock-value deepfakes to subtle enhancements that are harder to detect and easier to defend.