AI deepfakes pose growing threat to U.S. midterm elections, experts warn

AI-Generated Videos Now a Threat to Election Integrity, Experts Warn

In March, the National Republican Senatorial Committee released a video of Democrat James Talarico, a candidate for a U.S. Senate seat in Texas, reading his old tweets. The video showed Talarico standing before an American flag, his voice sounding authentic. The problem: it wasn't him. The video was artificially generated.

This moment signals a shift in how generative video technology can be weaponized in politics. Experts say generative AI has improved significantly since the 2024 election cycle, and the technology now poses real risks to voter confidence and election outcomes in this fall's midterms.

"It is easier now than it was last year and two years ago to produce and disseminate false information, particularly using generative AI," said Thessalia Merivaki, an associate professor at Georgetown University who studies elections and democracy.

Public Detection Skills Lag Behind Technology

More than half of Americans said they cannot confidently identify AI-generated content, according to a September 2025 Pew Research Center study. Deepfakes-videos or images manipulated using AI-are driving a surge in misinformation.

During the 2024 election, text-based misinformation dominated because AI-generated video and images weren't realistic enough. That has changed. In recent months, the Mike Collins campaign produced an AI video showing Democratic Senator Jon Ossoff of Georgia pledging loyalty to Chuck Schumer. A Virginia group opposed to congressional reapportionment released an ad featuring an AI-generated woman resembling Governor Alison Spanberger burning down a barn.

Loreben Tuquero, who covers AI and misinformation at PolitiFact, said the technology's rapid improvement means deepfakes now look convincing at first glance. "There are instances of AI that could lead voters to believe something is true," Tuquero said.

How Social Media Amplifies the Problem

Social media platforms prioritize emotional content that drives engagement, creating what Merivaki calls an "uneven playing field" where low-quality information spreads faster than authoritative sources.

Generating convincing videos or images now costs less and requires less expertise than ever before. If deployed strategically near elections, AI-generated content about candidates could confuse voters or suppress turnout. "Generative AI is very convincing," Merivaki said. "That is no longer the case" that deepfakes are easy to detect.

Spotting Fakes: What Voters Can Do

Tuquero offers practical steps for identifying deepfakes:

• AI-generated videos are typically short-around eight seconds-which can signal they're fake.
• Look closely at details like teeth and hands, which often appear strange or nonhuman in AI videos.
• Scroll with skepticism. If something looks off, seems untrue, or provokes strong emotion, dig deeper.

Tuquero warns against passive viewing. "Just be aware that something you're seeing-something you're watching and may like-could be fake, and develop steps for yourself to check whether it's real."

One complication: public officials may claim real videos are AI-generated to avoid accountability. This is known as the "liar's dividend."

Detection Tools Have Limits

AI detection tools exist but aren't fully reliable. An August study found that state-of-the-art commercial detection tools show moderate to high success rates, but false positives still occur. Matthew Wright, principal investigator of the DeFake Project at Rochester Institute of Technology, said current commercial tools are "so-so."

The DeFake Project is developing an all-in-one tool for journalists and other professionals to identify AI-generated content across media types. "It's not necessarily about changing anyone's mind-it's about getting accurate information out there," Wright said.

States Move to Regulate Political Deepfakes

Twenty-nine states now have laws regulating deepfakes in political messaging, according to the National Conference of State Legislatures. Most require disclosures on media like political ads if they contain AI-generated imagery, but few outright ban deepfakes.

Maryland's General Assembly passed legislation that would authorize the state administrator of elections to take action against election misinformation online, including deepfakes. The bill awaits Governor Wes Moore's signature.

As lawmakers respond, Wright offered straightforward advice: rely on trusted news organizations. "If somebody posts a video from a source you've never heard of and you're not sure about it, you probably shouldn't trust it," he said. "But if it comes from a reliable source, you can feel confident that journalists are doing their best."