Voters Turn to AI Chatbots for 2026 Midterm Research
Voters are using AI chatbots to research their ballots as midterm voting was underway on Oct. 5, 2026, in the first midterm since AI use grew sharply. NPR described the trend in a story titled "People are using AI chatbots to help decide who to vote for."
Adam Johnson, 40, of Morgantown, West Virginia, spent hours talking with ChatGPT about his 2026 ballot, according to NPR. His account was one of several voter examples in the report.
About half of American adults said they use AI-powered chatbots, according to Pew Research Center data cited by NPR. About 42 percent of those users said they use chatbots to search for information, according to the same data.
Voters told NPR they turn to chatbots for down-ballot races, the state and local contests that often get less news coverage. They said they ask the tools to build tables comparing candidates in governors' races. They also said they use chatbots to check viral claims.
ChatGPT is a chatbot built on a large language model, a type of AI trained on large amounts of text to generate answers, and is owned by OpenAI. Claude is owned by Anthropic. Gemini is owned by Google. OpenAI said it continues to "monitor bias in our models to keep ChatGPT's responses politically neutral."
Large language models are becoming a common source for election information, ahead of social media and search, according to the New York Times. Research on voters using AI ahead of the 2026 midterms found chatbots are sometimes helpful and sometimes misleading, USA Today reported.
The Brennan Center for Justice tested how well chatbots correct false claims about elections when asked, including by users who said they distrust election processes or results, according to the Brennan Center for Justice. Separate research found clear factual errors by chatbots have fallen ahead of the 2026 midterms, but incomplete answers are still a main concern, researchers said.
An expert told NPR that checking one chatbot against a different one is probably better than asking the same chatbot to fact-check itself, but it is not foolproof. It is not clear whether AI will become a voting issue in the November 2026 election.
The broader context here is familiar to people who run campaigns and elections. Down-ballot races often get less coverage, but voters still must decide them. A table or short summary can feel like enough when there is little else to read.
In my view, looking toward November, the concern is less about one wrong answer than about scale. If many voters ask similar questions and get similar incomplete summaries, mistakes or missing details can spread quietly. That puts a premium on clear, citable voter information that both people and chatbots can use.
