Researchers found some voters are using AI to understand politics ahead of the midterms, finding that the chatbots are sometimes useful, while other times they’re misleading.
The unlikely alliance pushing back on AI
What brought Sanders and Bannon together? Growing concerns that AI is moving too fast.
Creating recipes, finding news, generating images, answering research questions, writing emails, helping with job applications: the list of what people are using AI for goes on and on. But researchers suggest that the 2026 midterms are inspiring another use for the bots: understanding election information.
Recent AI conversations raising alarms that the technology could doom humanity have made their way into American politics in the past week. Lawmakers on Capitol Hill are considering regulations, while President Donald Trump is against reining in the growing technology. AI leaders came together to ask the government to slow development, and the issue even brought together political opposites, democratic socialist Sen. Bernie Sanders and conservative Trump ally Steve Bannon, at a Washington, D.C., event.
All this comes as the November midterm elections are just over six weeks away – and people are using the technology to learn more about the issues and candidates, according to researchers.
“We started hearing from friends around us that they have been using chatbots in order to obtain information about elections … so we started thinking about, ‘How does a chatbot that knows some information about you potentially change the information that it gives you with regards to elections?'” Massachusetts Institute of Technology Professor Chara Podimata told USA TODAY. She is conducting the research with political misinformation specialist Adam Berinsky and politics professor Charles Stewart.
How prevalent is AI use for election information?
About 10% of Americans report using chatbots to get their news, according to the Institute for Strategic Dialogue.
“While recent surveys show only a small proportion of American voters knowingly turn to AI chatbots for information about politics and elections, this share is likely to grow in the coming years,” Valeria de la Fuente, Max Read and Peter Benzoni wrote in a Sept. 3 report for the Institute for Strategic Dialogue.
A voter poll conducted by the Rainey Center, shared in June 2026, revealed that 9% of voters used AI for political information. The poll also reported that voters said they would use it for these election-related topics: 36% to understand candidate positions; 30% to compare candidates; 33% for learning how to vote; 45% to fact-check claims; and 44% to summarizing news/political topics.
The Rainey Center’s release added that while only 9% of people actively used AI for political information, a larger percentage of people “are already encountering AI-shaped content through Google, TikTok, and social search without knowing it.”
How accurate is AI about election information?
AI does a fair job at pushing back against election disinformation, according to researchers from the Brennan Center for Justice at the New York University School of Law.
However, the researchers, who published their work in August 2026, also noted that the chatbots easily generated “misleading election-related images and videos,” and that when those creations were fed to other chatbots, the AI systems were largely unsuccessful in identifying the images or videos as fake.
“This, along with other findings from our study, suggests that in an ongoing election with little authoritative information for models to rely on, AI may not be as helpful in pushing back against unsubstantiated rumors and misinformation,” the authors wrote.
In Podimata’s work at MIT, researchers created an ongoing audit to track how popular chatbots – including ChatGPT, Gemini, Grok and Claude – are responding to questions about the 2026 midterms. In the group’s original research leading up to the 2024 presidential election, they found that chatbots changed answers for the same question over time – often based on small differences in questions like “I’m a woman” or “I’m a man,” and those changes were not always in line with public announcements or election developments.
They continued the research ahead of the midterms to see how the chatbots would perform with lesser-known names (compared with Kamala Harris, Donald Trump and Joe Biden in 2024) of candidates in congressional races across the country.
“I want the public to know that the type of information that each of us gets may be different from what our friend gets for exactly the same question,” Podimata told USA TODAY.
AI is effective in shifting a voter’s opinion about presidential candidates or proposed policy, according to separate research from Cornell University’s College of Computing and Information Sciences released in December 2025.
A science writer for the university who reviewed the research, Patricia Waldron, called chatbots “quite effective at political persuasion” and said that the “potential for artificial intelligence to affect election results is a major public concern.”
How can chatbots help voters?
MIT’s Podimata told USA TODAY that she doesn’t “want to prescribe for anyone how to use their AI or tell them to stop doing it,” noting that AI in this use can “create even more engaged citizens.”
“It can maybe help you decide what ballot measure to vote for, which is something that in the past may have been very hard for people to do,” Podimata said. “Now they can converse with a chatbot and obtain this information fast.”
The Brennan Center research noted that chatbots “could play a constructive role in fighting election disinformation.”