Background: Generative artificial intelligence (GenAI) is becoming increasingly present in educational and everyday digital contexts, raising questions concerning what pupils should learn about these technologies and how such learning should be addressed in primary education. Existing research has largely focused on the adoption and use of AI tools in education. At the same time, less attention has been given to how teachers interpret the educational significance of GenAI and to the pedagogical expectations they construct for pupils’ AI-related learning. This study, therefore, explores how primary school teachers reason about what, why, and whether pupils should learn about GenAI in primary education.

Methods: The study adopted a qualitative, interpretive design based on semi-structured interviews with 17 primary school teachers from diverse educational contexts. Data were analysed using reflexive thematic analysis (Braun & Clarke, 2019, 2021). AI literacy, a sociomaterial perspective, and an ecological understanding of teacher agency informed the analysis.

Findings: The findings suggest that teachers’ pedagogical judgements regarding GenAI emerged through ongoing interpretation of changing educational and technological conditions. Teachers’ reasoning was shaped particularly by pupils’ existing engagement with GenAI, limited institutional guidance, concerns regarding dependency and critical thinking, and broader questions concerning the role of primary education in increasingly AI-mediated environments. Participants emphasised the importance of developing pupils’ critical understanding of how GenAI functions, recognising its limitations, and promoting responsible and reflective engagement. At the same time, teachers negotiated tensions between preparing pupils to engage with emerging technologies and protecting foundational learning processes such as reading, writing, independent thinking, and cognitive effort. The findings further demonstrate that teachers developed different pedagogical responses to GenAI depending on how they interpreted pupils’ needs, educational responsibility, and the purposes of primary education.

Conclusion: The study contributes by shifting analytical attention from questions of AI adoption toward teachers’ professional reasoning and pedagogical judgement and toward teacher agency in relation to GenAI. It further demonstrates how AI literacy is interpreted and constructed in practice rather than implemented as a fixed set of competencies. By combining a sociomaterial perspective with an ecological understanding of teacher agency, the study highlights how educational responses to GenAI emerge through interactions among technological developments, institutional conditions, pupils’ everyday practices, and teachers’ situated professional judgements within primary education.