AI Text Agents Reached 35,000 Voters in This Midterm Cycle

Campaigns across the country are deploying artificial intelligence text agents to contact voters during the 2026 midterm elections. These tools allow political teams to hold conversations with far more people than human staff can reach in a single day.

Ace Parsi, a Democratic candidate in West Virginia, uses the technology despite his personal skepticism about AI. His team sends texts to voters in his district ahead of the November election. Parsi told CBS News that the campaign views the bots as a way to make constituents feel heard. He noted that the strategy centers on ensuring voters feel seen and listened to.

The agents are powered by large language models, similar to the technology behind ChatGPT and Gemini. Campaigns load these models with approved materials such as speeches and press releases to limit what the bot discusses. Tom Carroll, founder of Convos.ai, said the agents only discuss information provided by the campaign. This approach aims to prevent the conversation from straying onto unapproved topics.

Vector Political, a Democratic firm, has sent over 3 million text messages this cycle. Those messages generated more than 35,000 individual conversations. Marty Santalucia, a partner at Vector, said the firm trains its agents to use deep canvassing techniques. This method focuses on extended, empathetic listening to understand voter concerns.

Research indicates the approach influences voter behavior. A study in the journal Nature tested chatbots in the 2024 U.S. presidential race. The experiment included 2,306 citizens. The study found that short AI conversations shifted voter views more than political ads typically do. The effect was strongest among voters who initially supported the opposing candidate.

However, the technology presents challenges. A study in the journal Science found that AI claims become less accurate when bots push harder for a specific viewpoint. Privacy questions also arise because campaigns receive full transcripts of the interactions. Most states do not require campaigns to disclose that voters are texting with an AI.

Parsi’s campaign sent 23,000 texts. Approximately 400 recipients, or 1.7 percent, engaged in a back-and-forth conversation with the agent. Monica Biddix, Parsi’s campaign manager, said the team receives every chat along with alignment scores. These scores range from 0 to 100, indicating how strongly a voter supports the candidate.

Parsi personally called every voter who expressed concern about the tool. He said he focused on people who disagreed with him. This practice contrasts with his view of how many elected officials communicate with their base.