AI Texting Bots Reach Over 3 Million Voters in Midterms
Campaigns in 2026 are using large language models to conduct thousands of text conversations daily, a shift that raises new privacy questions and offers a faster way to engage than traditional door-knocking.
By Dominic Foster ·
Voters across the United States are increasingly receiving text messages from artificial intelligence systems rather than human campaign staffers. In the 2026 midterm cycle, a growing number of political teams are deploying chatbots to conduct thousands of conversations in a single day. This method allows campaigns to reach a larger pool of constituents than volunteer canvassers can manage on the ground, while simultaneously collecting data on voter preferences and concerns.
The use of these tools is not new, but it has accelerated. In December 2025, a study published in the journal Nature examined chatbots used in the 2024 U.S. presidential election, the 2025 Canadian federal election, and the 2025 Polish presidential election. The U.S. portion of that study involved 2,306 citizens. The research found that short conversations with AI agents shifted voter views more than traditional political advertisements. Specifically, pro-Harris chatbots moved Trump supporters nearly 4 points on a 100-point scale toward Harris, an effect roughly four times larger than that of pro-Democratic video ads in previous cycles.
Despite the data, the technology is moving faster than the regulatory frameworks governing it. A companion study published in the journal Science tested 19 large language models across more than 91,000 conversations. The study found that as chatbots pushed harder to persuade, the accuracy of their claims decreased. Furthermore, most states do not currently require campaigns to inform voters that they are texting with an AI agent. In most jurisdictions, voters can receive full transcripts of their conversations, creating new privacy considerations for those who share personal details.
Campaigns are taking different approaches to mitigate these risks. Ace Parsi, a Democrat running in West Virginia, stated that his campaign tells voters at the start of each conversation that they are texting with a chatbot. His campaign sent 23,000 texts, and approximately 400 recipients, or 1.7 percent, engaged in a back-and-forth dialogue. Parsi said his team uses the data to identify voters who expressed concern about the tool and then follows up with human phone bankers. He noted that he personally called voters whom the AI ranked as strongly disagreeing with his campaign.
Marty Santalucia, a partner at the Democratic firm Vector Political, said his company has sent over 3 million text messages this cycle. These messages led to more than 35,000 individual conversations. Santalucia said the firm trains its AI agents to mimic the techniques of deep canvassing, a method based on extended, empathetic listening. The agents are instructed to listen to the voter's personal experiences and relate them to the campaign's goals. Santalucia said that among people who reply, 10 percent to 20 percent remain in the conversation for at least an hour.
Vendors like Convos.ai provide the technology by loading campaign-approved material into the system. Tom Carroll, the founder of Convos.ai, said the agents only discuss information provided by the campaign, such as speeches and press releases, to reduce the chance of errors. The system rates each voter from 0 to 100 based on their alignment with the candidate. A score of 0 indicates a strong detractor, while 100 indicates a supporter motivated to work on the campaign's behalf. Campaigns use these scores to decide whom to contact further and whom to remove from their lists.
Jessica Alter, co-founder of the Democratic-aligned group Tech for Campaigns, argued that chatbots can reach voters who decline to speak with human door-knockers. Alter said that if voters express a preference for chatting with a bot over talking to a stranger at their door, campaigns should listen to that preference. The technology allows for a level of engagement that is difficult to achieve with limited staff resources.
The next major test for these tools is the November general election. Campaigns are currently using the data collected from these AI-driven text exchanges to refine their messaging and target specific precincts. As the election approaches, the number of voters interacting with these digital assistants is expected to rise, making the accuracy and privacy of these conversations a central issue in the 2026 midterm race.
