At the National Conference on Citizenship, we have tracked America’s civic health for the last 20 years — and today, understanding civic health means understanding the impact artificial intelligence has on civic engagement. This election season, we focused this research on two questions. How many voters are turning to chatbots for information? And what do chatbots say when asked the questions voters are likely to ask?

To answer the first question, we worked with researchers at The Carter Center and the polling firm SSRS. Twice in October, we asked a national sample of about 1,030 U.S. adults several questions, including whether they had used an AI chat program to learn about candidates for Senate, House, or governor as they decided how to vote. In the first days of the month, 14% said yes. Two weeks later, 25% did. Only 22% trusted AI a great deal or quite a bit to give unbiased information about candidates.