States use AI systems to determine eligibility, verify identity, and detect alleged fraud. States use AI chatbots to answer questions about UI. Nevada even plans to use Google AI to write decisions in unemployment appeals.

These systems generally take information from applicants or recipients, process it, compare it against other sources of data, and send out requests for information and notices. The system may either make a decision outright or prepare a decision for an agency employee to make. When these systems do not work properly, people who need and qualify for UI go without. 

While AI chatbots do not make decisions about eligibility, they shape people’s understanding of UI eligibility requirements, application steps, and how to comply with the program rules. Generative AI chatbots write their own answers to questions, rather than using preprogrammed responses that were written by agency staff. Generative AI tools are known to provide wrong answers or invent information. For example, New York City’s government created a chatbot that gave businesses wrong information about the law. It was used for more than two years before city officials deactivated it.