Generative artificial intelligence (AI) offers new opportunities for history education, such as the ability to chat with historical figures. However, little is known about pupils' interaction with AI applications such as ChatGPT. A qualitative case study was conducted to explore how pupils (n = 21, year nine, fourteen years old) interacted with ChatGPT while completing a history-learning task that involved interviewing the chatbot of a historical figure. The resulting chat protocols were analyzed using grounded theory, focusing on a sociocultural approach to history education and concepts from human--machine communication (HMC). The findings suggest that pupils' interaction with AI can be seen as a kind of human--machine communication involving specific types of history-related prompts from humans and particular responses from AI. Based on these findings, the article discusses implications for further research and practice.