Voters Are Turning to ChatGPT for Election Advice: One User Warns the AI Is Just a ‘People Pleaser’

As the US midterm elections draw closer, a growing number of voters are turning to AI chatbots such as ChatGPT to help research their ballots and decide who to support, with experts warning that the technology could reinforce what users already believe.

About one in five voters said they consult AI chatbots for election news and voting, according to a September survey by the American Association of Political Consultants Foundation. Democratic and Republican voters reported using the technology at similar rates.

Voters are using chatbots in a variety of ways, from researching lesser-known candidates and comparing their positions to checking viral claims and creating tools that help others find information about their ballots. The trend comes as some voters face thinner local media coverage of down-ballot races, giving AI another role in how they gather political information.

ChatGPT May Tell Voters What They Want to Hear

Experts warn that the same qualities that make chatbots useful could also make them subtly persuasive by reinforcing what users already believe.

'It might select some things that would reinforce and persuade you even more towards the way that you were leaning already,' Rafael Batista, a postdoctoral fellow at Johns Hopkins University who studies the societal and behavioural impacts of AI, told NPR. 'So you leave more confident, without necessarily learning more about the world.'

Batista also pointed out that AI companies do not disclose exactly how their commercial models are trained or how they decide which sources to use, making it difficult to know what biases could shape their answers.

An OpenAI spokesperson directed NPR to the company's election information and safeguards page, which says it continues to monitor bias in its models to keep ChatGPT's responses politically neutral.

Adam Johnson, a 40-year-old from West Virginia, is among the voters using ChatGPT to prepare for the midterms. He spent hours discussing his ballot with the chatbot, asking it to list the races and then digging deeper into the candidates and issues that matter to him.

Johnson usually votes Republican unless a candidate is part of the MAGA movement, in which case he may vote for a Democrat or third-party candidate. While discussing the Senate race, he also asked ChatGPT to consider his views on immigration and refugees.

The chatbot told him it could discuss his guiding principles and compare candidates according to criteria he selected, while saying it should not manipulate his political views or tell him how to vote.

Johnson is nevertheless aware of what he sees as the chatbot's tendency to agree with users.

'ChatGPT is just kind of a people pleaser,' he told NPR. 'Whatever you lean towards, it will speak positively about it.'

Voters Are Using AI in Different Ways

Johnson is not alone in turning to AI for election research. Lisa Veldran, a retired Madison, Wisconsin, City Council staffer, used Google's Gemini to create a comparative table of candidates during the state's gubernatorial primary.

Veldran said she sees chatbots as a research tool rather than something that should determine her vote, while also questioning who is behind the technology and what their ultimate goals might be.

Lakshmi Iyer, a 50-year-old author in Pennsylvania, has also used AI for political purposes, although not to research her own ballot. She used Claude to create Ballot Lookup, a volunteer-built, non-partisan website designed to help Pennsylvania voters find out who is on their ballot and where candidates stand on certain issues.

Iyer believes human judgement remains essential when using AI. She also recognises the possibility that chatbots could infer a user's political leanings and push them further in that direction.

To keep her own bias in check, she said she prompted Claude to use statements from candidates' own speeches or official websites rather than relying solely on the chatbot's interpretation.

Other voters are turning to AI partly because local media coverage has declined, leaving fewer sources for information about lesser-known candidates. Some are also aware that chatbots can produce false or misleading information and attempt to cross-check answers using other AI systems.

Batista cautioned, however, that he has not seen evidence showing that using another chatbot to fact-check the first one is necessarily effective.

AI Could Reinforce Political Biases

The concern, then, is not simply that AI might give voters an incorrect fact. A more subtle concern is that a chatbot could present information in a way that confirms what someone already believes, potentially making them more confident without giving them a fuller understanding of the issue.

That matters because conversational AI can feel more like a discussion than a traditional search engine. Users can explain their beliefs, challenge responses and ask follow-up questions. Batista's concern is that information already available to a chatbot could influence how it responds.

Batista suggested that people using chatbots to research their vote could frame prompts without revealing their personal preferences, and could use incognito mode to limit the amount of prior information available to the chatbot.

For voters, the challenge is therefore not simply deciding whether AI is useful. It is knowing when a tool designed to provide political information might nevertheless reinforce an existing point of view.

Johnson remains in the process of researching his ballot, including local county commission races. But after using ChatGPT to investigate the other contests, he said he felt prepared to vote.

That confidence may be part of what makes the technology so appealing — but Batista's warning is that feeling more certain does not necessarily mean a voter has learned more about the world.