Adam Johnson, a 40-year-old graphic designer from Morgantown, West Virginia, spent hours talking through his 2026 ballot with ChatGPT, asking it to list his races, compare candidates and measure them against his stated political principles.
Roughly half of U.S. adults now use AI chatbots, according to Pew Research Center. Some of those users are turning those tools toward ballot research, a task the systems were not specifically built to handle.
How Voters Are Putting Chatbots to Work
Voters are using AI for everything from candidate comparison tables to volunteer-built ballot guides, particularly in races where local coverage is sparse.
Johnson told ChatGPT he typically votes Republican but avoids candidates associated with the MAGA movement. The chatbot responded that it could evaluate candidates according to criteria he selected but would not direct his vote, according to NPR.
Lisa Veldran, a retired Madison, Wisconsin, city-council staffer, asked Google’s Gemini to generate a comparison table after Wisconsin Lt. Gov. Sara Rodriguez dropped out of the governor’s race, according to NPR.
For down-ballot races with little coverage, Nashville voter Tyler Black turned to chatbots as a starting point. Pennsylvania author Lakshmi Iyer used Claude to build Ballot Lookup, a volunteer-run nonpartisan site, instructing the model to draw from candidates’ own speeches and official websites rather than party labels, according to NPR.
The Risk Hiding Inside the Helpfulness
A chatbot that agrees with you is not the same as a chatbot that informs you.
“It might select some things that would reinforce and persuade you even more towards the way that you were leaning already,” said Rafael Batista, a Johns Hopkins University fellow who studies how AI shapes people’s experience of the world, according to NPR.
Batista’s broader warning is that personalized answers can increase your confidence without increasing your understanding. That agreeable quality operates subtly, even during what feels like a neutral research session.
Chatbots can also hallucinate, producing false or fabricated information. Black told NPR he cross-checks one chatbot’s answers against another’s, but Batista said there is no evidence that method reliably catches errors.
“ChatGPT is just kind of a people pleaser,” Johnson told NPR.
Commercial AI companies do not fully disclose how their models are trained or how they rank political information, according to NPR. That makes independent bias assessment difficult.
What OpenAI Says It Is Doing
OpenAI has published election safeguards, though those are the company’s own stated commitments rather than independently verified outcomes.
OpenAI says it monitors ChatGPT for political bias and aims to keep responses neutral. The company has partnered with Democracy Works to provide voting logistics, and its stated policies prohibit election interference, voter demobilization and scaled campaign messaging, according to OpenAI’s election safeguards page.
How to Use These Tools More Carefully
A few prompt habits can reduce, though not eliminate, the risk of personalized persuasion.
Batista recommends framing prompts without disclosing your preferences, avoiding leading questions and using incognito mode to limit how much prior information shapes the response, according to TPR. Asking the chatbot to separate documented positions from inferred ones is also worthwhile.
Then verify what it tells you against official campaign websites, legislative records and reputable reporting. Cross-checking chatbots against each other can surface discrepancies, but it does not replace primary sources, according to NPR. For a broader set of vetted tools, AI-powered websites can supplement your research beyond what chatbots alone provide.
The chatbot can organize your ballot. Only you can verify what it tells you.
