A new Scripps News/Talker Research poll of 1,000 US registered voters finds that 43% say they’ve seen something in a political ad this campaign that they thought was real — only to later find out it was AI-generated. Another 30% say they’ve seen something they still aren’t sure about. And the demand underneath the numbers is unambiguous: more than three-quarters of voters want rules around how AI is used in politics, including 77% of Democrats and 79% of Republicans.
🔍 THE BOTTOM LINE
This campaign cycle already had the receipts — researchers tracked about $80 million spent on nearly 170 ads using AI, with only 31% of those ads disclosing it. What the poll adds is the viewer’s side of the ledger: the audience knows it’s being fooled, can’t reliably tell which ads are synthetic, and wants rules regardless of party. Disclosure laws exist in 26 states for deepfakes, but the researchers found ads in law-covered states disclose at 29% versus 32% elsewhere — the laws, as written, are not bending behaviour.
What the Numbers Say
The poll, commissioned by Scripps News and conducted by Talker Research, is a random double-opt-in online survey of 1,000 US registered voters, collected September 30–October 1, 2026, with a margin of error of ±2.2 points at 95% confidence. The headline: 43% of registered voters say they saw something they thought was real in a political ad and later found it was AI-generated, with another 30% saying they saw something they weren’t sure was real or AI. Nearly 80% say they see content they believe was AI-generated at least once a day.
The partisan-use numbers track the research: 56% of respondents said they’d seen Republicans using AI ads, versus 49% for Democrats. That matches the Wesleyan Media Project’s finding that Republican candidates or pro-Republican groups were behind 80% of AI-suspected ads and 83% of the spending — but, critically, the desire for rules crosses party lines almost evenly in the poll. Concern about AI also runs deeper than ads: 61% worry about AI’s impact on their jobs, 83% on government institutions, 81% on financial institutions, 57% on their physical safety.
What’s Actually Running on Screens
Some of the cycle’s most-discussed ads, according to Scripps’ report and prior coverage, featured AI-generated candidates — Sen. Jon Ossoff in a beauty salon, James Talarico reciting old quotes out of context, Sherrod Brown overstaying his welcome at a birthday party. The Wesleyan researchers catalogued deepfakes of Trump, Pelosi, Obama, Harris and Ocasio-Cortez (a favourite of Republican advertisers, appearing in at least five ads), a fake Dr. Fauci running around a state fair with a giant syringe, and a Louisiana Senate candidate driving a school bus full of undocumented immigrants.
The subtler use is harder for voters to clock: AI used to enhance visuals rather than invent them doesn’t necessarily look different from a pre-AI ad — “ads that use AI to enhance visuals don’t necessarily look different from ads that were created in the pre-AI era,” the researchers note. One Oklahoma ad’s disclaimer read “You guessed it! AI was definitely used to generate these silly video clips.” That mix — hyperreal fakes adjacent to joke-level disclosure — is exactly the environment where 43% of people report being fooled.
The Regulation Gap the Poll Exposes
Twenty-six states regulate political deepfakes, per the National Conference of State Legislatures, requiring disclosure or banning deceptive synthetic video in the weeks before an election. But Wesleyan’s data shows the disclaimers aren’t following the laws: 32% disclosure in states without AI-ad laws, 29% in states with them — statistically indistinguishable. At the federal level, the REAL Political Advertisements Act, introduced in 2023 by Sen. Amy Klobuchar and Rep. Yvette Clark, never reached a vote in either chamber, and no federal requirement exists this cycle. Minnesota bans political deepfakes generally, yet a synthetic video of Democratic Senate candidate Peggy Flanagan aired in May 2026 anyway — enforcement turns on whether the fake was “so realistic that a reasonable person would believe it.”
The New Zealand Mirror
The pattern is not American-only. New Zealand’s Parliament gave a deepfake-focused bill a unanimous first reading this year, and US states are running their own patchwork experiment that reads here as a warning: laws target the hyperreal deepfake and miss the subtle enhancement, which is precisely the category voters report failing to detect. Most Americans told AP-NORC that AI is moving too fast — that survey’s unease and this poll’s deception numbers point the same direction. New Zealand’s next election cycle will be the first where local campaigns can cheaply generate synthetic speakers; the polling lesson from the US is that “it’s obviously fake” is an assumption voters themselves no longer make, and disclosure rates won’t fix that without enforcement.
The AI-in-politics rules race now has a measured baseline: the audience wants rules, the ads don’t disclose, and the campaigns know it.
❓ FAQ
What did the Scripps poll actually find? In a survey of 1,000 US registered voters (Sept 30–Oct 1, 2026, ±2.2 points), 43% said they saw something in a political ad they thought was real and later learned it was AI-generated; 30% said they weren’t sure whether something they saw was real or AI. About 80% say they see believed-AI content daily.
How much AI political advertising is running? The Wesleyan Media Project tracked roughly $80 million spent on nearly 170 unique ads using AI, with 69% carrying no AI disclosure, and 80% of the ads coming from Republican candidates or Republican-leaning groups.
Do any laws require AI disclosure in political ads? Twenty-six US states have laws regulating political deepfakes, but federal legislation (the REAL Political Advertisements Act) has never reached a vote. Wesleyan found disclosure rates are effectively the same in states with and without AI-ad laws.
What about New Zealand? New Zealand’s Parliament gave a deepfake-focused bill a unanimous first reading this year. The US experience — laws that cover hyperreal deepfakes but miss subtle AI enhancement, low disclosure, voters unable to tell — is the closest thing to a preview of what local campaigns will face next election.
🔍 THE BOTTOM LINE
Polls about AI usually measure worry; this one measures experience — nearly half the audience says the technology already fooled them once this campaign, before Election Day. Disclosure mandates were the policy answer of 2024-2026 and the data shows they are not being heeded where they exist. The question the midterms will settle is whether campaigns learn that synthetic persuasion carries a backlash cost, or whether 43% quietly becomes the price of doing business.
📰 Sources
- Scripps News/Talker Research poll — 1,000 US registered voters, Sept 30–Oct 1 2026 (reported 14 October 2026)
- Wesleyan Media Project — AI in political advertising release (September 2026)
- The Conversation via The Independent — $80M tracked on almost 170 AI-generated midterm ads (October 2026)
- NBC News via AOL syndication — AI-generated political ads stoke fears of voter deception
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
