Midterm voters are turning to AI chatbots to check facts, compare candidates, and build their own tools for looking up races. A new wave of political research is unfolding through tools like ChatGPT, Gemini, and Claude, with voters using them to fill gaps in coverage for down-ballot races, verify viral claims, and create nonpartisan lookup sites.
The shift is visible across several states and races ahead of November’s midterms. Here is how it is playing out.
Adam Johnson’s Ballot in Morgantown
Adam Johnson, 40, used ChatGPT to research his ballot for the midterm election in Morgantown, West Virginia. ChatGPT is a chatbot powered by a large language model (LLM) owned by OpenAI, an AI research company.
Johnson is a graphic designer studying to become an illustrator. He uses ChatGPT for meal planning, job hunting, and household tasks. It has now become his research assistant for voting too.
About half of American adults report using AI-powered chatbots, according to Pew Research Center. About 42 percent of users said they use AI to search for information, according to Pew Research Center.
The Governor’s Race in Wisconsin
Lisa Veldran, 65, used Gemini to research the primary election for governor in Wisconsin. She faced a choice between Francesca Hong and David Crowley after Sara Rodriguez, the lieutenant governor, dropped out of the race.
Veldran said chatbots can be research tools but cautioned about who pays to build them and their ultimate goals. That caution is central to how voters are approaching these tools.
Claude Builds a Lookup Tool
Lakshmi Iyer, 50, is an author and part of They See Blue, an Indian American grassroots Democratic organization. Iyer used Claude to create Ballot Lookup, a Pennsylvania voter information website.
The site is described as an “independent, volunteer-built, nonpartisan lookup.” Iyer’s project shows how voters are not just consuming AI research but building it themselves to serve their communities.
What ChatGPT Found on Candidates
Johnson said he usually votes Republican, except when candidates are part of the Make America Great Again movement, in which case he votes Democrat or third party. When asked about immigration, ChatGPT noted Senator Shelley Moore Capito supported stronger border enforcement, ICE/CBP funding, the Laken Riley Act, and Trump administration border policies.
Rachel Fetty Anderson had less detailed immigration material publicly available, so ChatGPT avoided assigning her generic Democratic positions. The example shows how chatbots handle candidates with thin records compared to those with thicker ones.
Johnson called ChatGPT “a people pleaser” who speaks positively about whatever the user leans toward. That characterization matters because it points to a pattern in how these tools shape what voters see.
The Concern From Johns Hopkins
Rafael Batista, a fellow at Johns Hopkins University, studies how AI shapes how people experience the world. He raised a warning about the technology’s effect on voters.
“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.”
That concern is not abstract. It applies directly to how voters like Johnson use ChatGPT to research candidates and policies. The tool may confirm existing beliefs rather than expand them.
OpenAI’s Response to NPR
A spokesperson for OpenAI directed NPR to its election information and safeguards page, which states the company monitors bias to keep ChatGPT’s responses politically neutral. The company’s position is that it watches for bias and tries to keep responses balanced.
Anthropic, owner of Gemini, did not respond to a request for comment. That silence adds to Veldran’s personal concern about who builds these tools and what their goals are.
Comparing the Tools
| Tool | Owner | Voter Use |
|---|---|---|
| ChatGPT | OpenAI | Ballot research, fact-checking |
| Gemini | Anthropic | Primary research in Wisconsin |
| Claude | Anthropic | Ballot Lookup tool |
| Ballot Lookup | Volunteer-built | Nonpartisan voter info |
The Bottom Line on Voting
Voters are increasingly using AI chatbots to fact-check and compare candidates. The tools are becoming a regular part of how people prepare for elections.
The question is whether that is a good thing. Batista’s warning suggests the technology could reinforce existing beliefs rather than challenge them. Voters need to be aware of that risk as they lean on these tools.
The evidence so far is mixed. Johnson found useful information on candidates. Iyer built a valuable resource for voters. But the potential for confirmation bias is real.
General advice from experts is that voters should use these tools to learn but use them critically. Check the sources behind what the bot says. Question whether the answers match what you already believed.
The technology is changing how people engage with politics. The smart voter will watch it closely.
Where the paper stands
The paper backs voters using AI chatbots to check facts and build nonpartisan lookup tools, and is against any licensing regime that would freeze today’s leaders in place and lock out whoever would have challenged them. The tools described here — ChatGPT, Gemini, and Claude — are being used by ordinary citizens for exactly the kind of civic work the paper supports: fact-checking, candidate comparison, and independent lookup sites.
What makes this story fit the paper’s position is the scale of the shift: voters are not waiting for big firms to provide them with election information. They are building it themselves, from lookup sites like Ballot Lookup to individual ballot research in West Virginia. This is not a top-down market; it is a bottom-up one, and the paper wants to keep it that way.
The concern from Johns Hopkins is real, and the paper shares it: the risk that these tools could confirm existing beliefs rather than expand them. But that concern does not justify regulation that would freeze today’s leaders in place and lock out whoever would have challenged them. The paper’s position is that narrow rules against direct harm — such as requiring companies to disclose safety failures they hid — are enough. Broad rules hand the market to the incumbents, and the paper opposes them.
Source material: “Voters are increasingly using AI chatbots to fact-check and compare candidates,” NPR.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.