Forty-year-old Adam Johnson sat on his porch in Morgantown, West Virginia one September evening and worked through his midterm ballot with ChatGPT, an artificial intelligence (AI) chatbot owned by OpenAI, according to an NPR report published Monday, Oct. 5.
Johnson – a graphic designer studying to become an illustrator – asked the chatbot to list the races on his ballot, then described his political principles to it, NPR reported. Johnson is one of a growing number of U.S. voters using AI chatbots to research their ballots ahead of the midterm elections.
About half of American adults report using AI-powered chatbots, and about 42% of those users said they use them to search for information, according to the Pew Research Center, as cited by NPR. About one in five voters said they consult AI chatbots for election news and voting, and Democratic and Republican voters are using bots at similar rates, according to a September survey conducted by the American Association of Political Consultants Foundation, a nonprofit focusing on political expression [1].
Chatbots Offer Convenience as Local News Coverage Shrinks
Several voters told NPR they rely on chatbots for down-ballot races with thinner news coverage, to build tables comparing candidates in governors' races and to verify or debunk viral claims. Some have used the tools to create resources for other voters.
Lisa Veldran, 65, a retired staffer who spent nearly four decades at the Madison, Wisconsin City Council, used Gemini this summer to answer specific questions about the primary election for governor. After Lt. Gov. Sara Rodriguez dropped out of the race, Veldran prompted Gemini to make a comparative table of candidates Francesca Hong and David Crowley.
Lakshmi Iyer, 50, an author in Exton, Pennsylvania, said she did not use a chatbot to research her own ballot but used Claude to create Ballot Lookup, described as an "independent, volunteer-built, nonpartisan lookup" to inform Pennsylvania voters about who is on their ballot and where candidates stand on issues. Tyler Black of Nashville, Tennessee, said he uses chatbots to learn about local candidates and often checks the information from one chatbot against another because chatbots sometimes hallucinate, a term for responses that are false, misleading or fabricated.
Experts Warn of Sycophancy and Hidden Bias in AI Models
Rafael Batista, a postdoctoral fellow at Johns Hopkins University who studies the societal and behavioral impacts of AI, told NPR that chatbots can reinforce existing political leanings and make users more confident without necessarily learning more about the world. "It might select some things that would reinforce and persuade you even more towards the way that you were leaning already," Batista said. "So you leave more confident, without necessarily learning more about the world."
Batista said it is hard to know what biases might shape chatbot answers because AI companies do not disclose how their commercial models are trained or how they choose which sources to draw on. OpenAI, which owns ChatGPT, directed NPR to its election information and safeguards page, which notes that the company continues to monitor bias in its models to keep ChatGPT's responses politically neutral. Anthropic, which owns Claude, and Gemini's owner Google did not respond to a request for comment, NPR reported.
Batista said knowing these models are agreeable and persuasive is not enough to protect users' minds from their influence. Independent researchers have repeatedly documented reliability problems in election-related chatbot responses, including a study whose authors said none of the models tested were able to correctly answer how many days were left before the 2024 general election [2].
Users Acknowledge Persuasion Risks and Adopt Guardrails
Johnson said he recognizes that AI-powered chatbots have sycophantic qualities. "ChatGPT is just kind of a people pleaser," he said. "Whatever you lean towards, it will speak positively about it." According to NPR, Johnson described himself as usually voting for Republicans unless those candidates are part of the Make America Great Again movement, in which case he votes for Democrats or a third party.
Iyer said "the human in the loop is absolutely essential" and that if chatbots "can infer your political leanings, they will push you towards what you are leaning to." To keep her own bias in check, Iyer said she prompted Claude to pick statements from candidates' own speeches or from their official websites.
Batista recommended that users frame prompts so their personal preferences are hidden and use incognito mode so the chatbot has less prior information about them. He said checking one chatbot's output against another is better than using the same model twice, though he said he has not seen evidence that the method is effective.
Centralized Control and Privacy Questions Remain
The NPR report noted that AI companies do not disclose how their commercial models are trained or which sources they draw on, leaving questions about central control over election information. Because a handful of companies operate the most widely used chatbots, the criteria those systems use to weigh sources are not visible to the voters who rely on them.
Critics of Big Tech have raised concerns about privacy, data collection and the potential for centralized AI systems to shape what users see. Privacy researchers have documented how AI assistants can collect intimate details from conversations, including health concerns and emotional struggles, without clear opt-out options [3]. In a November 2025 incident, hackers stole OpenAI user data through a breach at its analytics partner Mixpanel, compromising names, email addresses and user location information [4].
Some decentralization advocates recommend independent, privacy-focused AI alternatives, but the NPR report did not evaluate such tools. The report also did not describe regulatory action related to voter use of chatbots.
Outlook: Johnson Says AI Helped Him Prepare for November
Johnson said he is still researching and chatting with ChatGPT about his ballot, with county commission races still on his list to look into. But he said he feels AI helped him learn what he needed to know for November.
"I wouldn't feel too bad if I had to do the election today," Johnson said, according to NPR. The report did not state whether any election officials or regulators plan to address voter use of AI chatbots.
References
- People Are Asking A.I. Chat Bots To Help Them Decide For Midterm.... 2026-10-06T01:20:17Z. https://gagadaily.com/forums/topic/431455-people-are-asking-ai-chat-bots-to-help-them-decide-for-midterm-election/
- AI chatbots got questions about the 2024 election wrong 27% of the.... 2024-06-07T21:32:58Z. https://www.nbcnews.com/tech/tech-news/ai-chatbots-got-questions-2024-election-wrong-27-time-study-finds-rcna155640
- How Metas free AI assistant turns private conversations into profit - NaturalNews.com, May 11, 2025. by NaturalNews.com
- OpenAI data breach exposes risks of third party vendors in AI industry - NaturalNews.com, November 28, 2025. by NaturalNews.com
Explainer Infographic
Please contact us for more information.