Artificial intelligence has been woven into the everyday lives of Americans. As voters head to the polls on November 3 for a consequential midterm election, will AI change the way they get their information and ultimately influence how they vote?
Since the 2024 presidential election, AI use has grown significantly, according to a Pew Research Center poll, which found that half of U.S. adults use AI, with a quarter of them using it daily. Two years ago, only one-third of adults had used AI platforms.
Public skepticism has grown alongside that adoption. A CNN poll found that 82% of respondents are more concerned than excited about the rapid pace of AI development, and 71% want the government to regulate AI more.
As the debate about AI rages on and its use becomes more ubiquitous, this year’s midterms could be a major test of AI’s ability to influence election outcomes, UC Berkeley researchers say. While chatbots can be used to better educate voters about the issues, they can also generate and proliferate misinformation that makes it difficult for the electorate to separate fact from fiction. For researchers, the central question is whether voters can make informed decisions on their ballot or be swayed by AI-generated falsehoods to vote a certain way.
Misinformation proliferation
AI’s mainstream use has enabled politicians and the public to spread misinformation at an alarming rate with tools such as deepfakes, which are AI-generated content showing real people doing things they never did, according to the findings from the Wesleyan Media Project, which tracks political advertising spending and content. Lawmakers in 30 states banned AI-generated deepfakes in election communications in response.
But that hasn’t stopped political campaigns from using AI-generated ads. The Wesleyan Media Project found that the 2026 midterm cycle included 170 AI-generated political ads, representing $80 million worth of ad spending during. More than two-thirds of those ads did not disclose their use of AI.
Undisclosed ads like these can create alternate facts and blur the line between reality and falsehoods, researchers say. For UC Berkeley Political Science Professor Susan Hyde, the concern is whether voters can still make informed decisions when they can no longer trust the information they receive.
“Classical theorists of democracy focus on the necessity for citizens to be informed about their choices in order to make an educated choice about who to vote for,” Hyde said. “The other thing that’s interesting is the ease with which politicians can use propaganda-style communication and lie to their supporters directly.”
UC Berkeley Political Science Professor Gabriel Lenz argues that misinformation during an election cycle is not unique to AI, as politicians have always made false claims and spread misleading information in order to win their races. That type of misinformation can influence voters, he said, but issues that directly impact Americans — inflation, unemployment and the economy — still have a larger sway over election outcomes.
He pointed to the Vietnam War, when President Lyndon Johnson’s claim that peace was imminent produced a positive shift in polling, though only temporarily.
“We haven’t really seen [AI campaigning] fully tested out yet, but we may see it this election,” Lenz said. “But it’s also always been true that incumbent presidents and candidate campaigns more generally make false claims about performance and push those claims on voters, especially as Election Day approaches.”
Can AI help voters?
Still, many AI companies and researchers see an opportunity for chatbots to provide accurate, tailored and easily digestible information. Anthropic and OpenAI have both partnered with organizations such as Democracy Works and AP News to provide accurate election information.
Rather than integrate into a larger chatbot, some smaller platforms are created specifically for political explanations. VoteMate, for example, can explain elections, including local races, where information is often hardest to find.
Stanford Sociology PhD student Joe Mernyk has examined how voters respond to these tools. Mernyk and his colleagues published a study that found participants rated AI information from their chatbot, which was designed to provide nonpartisan information, as trustworthy, accurate and unbiased. They also found that people who used their tool reported greater intentions to vote, which they speculate is a result of greater access to voter information, boosting a voter’s confidence at the ballot box.
“The people who might benefit most from accessing traditional voter guides are also the least likely to access information of this kind,” Mernyk said. “Sometimes, when people don’t have good access to information, they’re not making the most informed choices about which candidates share their views on the issues that they care about, and sometimes it means that they just don’t vote.”
Mernyk said he thinks AI-generated voter guides could be more helpful and digestible than traditional voter guides because it’s interactive and tailored to what voters are seeking. Voters can ask AI about the issues they care about most and receive a neutral answer, rather than wading through a physical and static voter guide.
However, the jury is still out on whether AI will help or hurt voters during the midterms.
“I think the counterfactual you should always be thinking about with voter information is what are their current sources of information and will chatbots be a net improvement on that or a net negative,” Lenz said. “That's a really hard question to assess.”