Large language model-based chatbots can't effectively recognize users' motivation when they are hesitant about making healthy behavior changes, but they can support those who are committed to take action, say researchers.