AI bots text voters as campaigns seek personalized outreach

Convos can respond within 30 seconds, in any language, and some voters end up talking to the agent for hours. AI-powered campaign texting tools are turning mass outreach into something that can feel like a private exchange with a candidate, holding personalized conversations with thousands of voters at once and logging what those voters care about so campaigns can shape the next round of messages.

Many of Aaron Sheeks’ current clients at Akillion are running for political office, and the tools give campaigns “a trained AI employee” that can handle questions on police reform, education or tax changes. In most cases, a human still sends the first text and the AI steps in only after the recipient responds. The technology can make campaigns more interactive, more responsive and more personalized while helping them do more with less. Tom Carroll, chief executive of Convos: “the era of the long political text message is over.” Convos launched last year, worked with 10 political campaigns and is aiming for more than 100 this year, with about half that target already reached.

The legal framework around campaign texting still hinges on consent. Political robotexts to mobile phones generally require prior express consent under the Telephone Consumer Protection Act. Manually sent political texts can go out without prior consent, but campaigns must honor opt-out requests, including a reply of “stop.” Calls made with AI-generated voices count as “artificial” under the law, following an investigation into illegal robocalls to New Hampshire voters that used deepfake voice technology. In 2024, the FCC proposed the first federal rules specifically aimed at AI-generated robocalls and robotexts.

States have moved faster than Washington. The National Conference of State Legislatures counted 31 states as having enacted laws regulating deepfakes in political messaging as of June 23, 2026. Minnesota and Texas ban publication of political deepfakes within certain windows before an election, Maryland bars deceptive deepfakes year-round, and Colorado and Utah require metadata disclosures for some political deepfake media.

Public skepticism remains high. AP-NORC and USAFacts polling found about 4 in 10 Americans said AI would make it harder to find factual election information, while another 4 in 10 were unsure.