By Danielle Ngamegni
University of Maryland students expressed uncertainty about an artificial intelligence-powered chatbot designed to provide election information to voters for the 2026 midterm elections.
Cody Buntain, principal investigator of Voter Information from Official Local Election Trusted Sources, or VIOLETS, says the project is an experiment to assess whether providing local election procedural information improves people’s perceptions of election integrity.
While the chatbot aims to strengthen democracy, some students questioned whether they could trust the technology and its motives.
Aakriti Mishra, a senior neuroscience major, said she is skeptical of AI companies providing voters with inaccurate or misleading election information, but is more open to the development of VIOLETS.
“I guess trust it a little bit more, but I still don’t have complete trust in it because AI isn’t really known to be very accurate,” Mishra said.
Ronan O’Beirne, a junior history major, says he would have to determine whether the chatbot sources its information from what he considers reputable sources before deciding whether to trust it.
“It’s a lot of skepticism again about the accuracy of AI, and I wouldn’t be surprised if people were initially resistant to using it for any kind of election information because of that,” he said.
Do Won Kim, a graduate research assistant leading the VIOLETS project, says the chatbot uses retrieval-augmented generation, an AI framework that uses information from external sources to generate a response.
According to Global Elections and Information Security (GEIS), its databases first retrieve and gather information from the Maryland State Board of Elections and the Montgomery County Board of Elections. Users would submit queries on VIOLETS, then OpenAI’s GPT-4o-mini would generate a response.
Kim explained that looking at election or voting information across multiple government websites can become burdensome for voters.
“When we use [an] AI chatbot in a conversation, we kind of ask follow-up questions to understand the information that AI gives us and AI kind of summarizes and makes election information easier to understand,” she said.
VIOLETS also incorporates hallucination prevention measures, including citation requirements, website security blockers and human review to reduce the risk of the chatbot generating fabricated information.
Buntain says researchers are primarily trying to determine whether it is more valuable to spend money on creating efficient AI chatbots or making government websites more accessible to the public.
Fatoumata Diallo, a sophomore public health science major, said she’s open to using AI, but worries how the public would approach it.
“I think I would be open to use it … I want it to be in a final form before it gets released to the public because people really don’t fact-check AI. They really take it [at] face value,” Diallo said.
Featured Image: A statue of Testudo sits in the lobby of the Adele H. Stamp Student Union, near the voting location on Nov. 3, 2026. Photo by Paige Trendell.