<p>This study investigates the nature of dialogues between preservice teachers (PSTs) and a chatbot (ChatGPT) in solving mathematical tasks. It explores how the PSTs assessed the chatbot’s solutions and how they supported it in solving the tasks. The dialogues of 24 PSTs are analysed through the concepts of telling (providing information or direction), assessment moves (to elicit student understanding), and advancing moves (to encourage deeper thinking) in teaching. The PSTs predominantly employed assessment and advancing moves over telling and, in some cases, employed subtle strategies that led the chatbot to revise its solutions. Unique to this study is the description of how the function of assessment and advancing moves are jointly constructed in the interaction between participants, challenging the idea that these moves can be defined solely by their form or the initiator’s intent. Interacting with a chatbot simulating a student offers a novel approximation of questioning and guiding individual students and an opportunity for reflecting on mathematics tasks and mathematical discussions.</p>