Abstract
This paper examines how three Australian primary teachers used generative AI to design and implement rich mathematical tasks. Drawing on multiple data sources analysed using three theoretical lenses, findings indicate that while generative AI expanded planning possibilities, the richness of tasks depended on teacher enactment rather than AI output. Teachers with deeper mathematical knowledge for teaching crafted more intentional prompts, critically evaluated outputs, and adapted tasks responsively, positioning professional knowledge as decisive in whether AI-supported task design enhances or constrains meaningful mathematical learning.