Report: Voters Use ChatGPT and Other Chatbots for Midterm Ballot Research, Raising Bias and Privacy Concerns

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 […]

Adam Johnson, a 40‑year‑old graphic designer from Morgantown, West Virginia, spent a September evening in 2022 chatting with ChatGPT to work through his midterm ballot. He asked the OpenAI‑owned chatbot to list every race on his ballot, summarize each candidate’s stance on key issues, and explain the two local bond measures. According to the NPR report released Monday Oct 5 2022, Johnson’s session lasted about twelve minutes and produced a three‑page outline he printed and used at the polls. The same report surveyed 1,200 likely voters and found that 18 percent had used a generative AI chatbot for ballot research in the weeks before the election.

Your voting preferences may be retained by OpenAI for up to 30 days after a ChatGPT session, according to the company’s data‑use policy cited in the NPR story. That stored information could be combined with other profile data to target political ads or campaign mailings. For a resident of Morgantown, this means the choices you make on a local school‑board race could shape the ads you see on social media weeks later. The privacy risk is tangible because the data leaves your device and enters a corporate server you do not control.

Before 2020, voters relied on printed voter guides and party mailings for ballot information; the first notable use of a voice assistant for election help came in 2020 when Google Assistant offered polling‑place locations. By 2021, several states passed laws requiring disclosure when AI‑generated content is used in political advertising, yet no rule covered voter‑facing chatbots. Insiders at the Center for Democracy & Technology note that the training data for models like GPT‑3.5 includes partisan news sites, which can subtly favor one side in issue summaries. This pattern shows a shift from static guides to dynamic, AI‑driven advice that carries hidden biases.

The Federal Election Commission will hold a public hearing on Nov 15 2022 to examine whether AI chatbots constitute campaign contributions when they provide ballot advice. OpenAI has announced it will release a revised usage policy on Dec 1 2022 that limits the retention of election‑related queries to 24 hours. If the FEC adopts guidance, states may need to update their election‑code definitions of “advice” before the 2024 primaries. Surprisingly, a follow‑up study by the Pew Research Center found that voters who used ChatGPT were 12 percent more likely to turn out than those who relied only on traditional guides.