["Journal of Computer Assisted Learning, Volume 42, Issue 4, August 2026. ", "\nABSTRACT\n\nBackground\nTeachers often lack the time and structured guidance needed to address the pedagogical and ethical challenges involved in the integration of artificial intelligence in education (AIED). Although prior research identifies these challenges, empirically grounded frameworks that help teachers integrate AIED in responsible and pedagogically meaningful ways remain limited. Addressing this gap is essential to ensure AIED enhances student engagement and higher‐order thinking skills (HOTS) rather than undermining them.\n\n\nObjectives\nThis study examines how teachers' understanding and adoption of AI tools change after participating in a collaborative professional development course on ethical and constructive AI use, and how students perceive and engage with AI‐supported learning. Drawing on study findings and the ACAD meta‐theoretical framework, the study proposes design dimensions for planning AI‐supported learning.\n\n\nMethods\nThe study involved 45 teachers and 266 students from three countries. Data sources included pre‐ and post‐course teacher surveys on ethical and constructive AI use; student surveys on engagement and perceived AI affordances; and qualitative analysis of 12 teacher‐designed lesson plans examining student engagement levels (informed by the ICAP framework), development of HOTS, and AI roles in planned activities.\n\n\nResults and Conclusions\nResults show that effective AI adoption is not intuitive. Teachers grounded in critical thinking pedagogy and collaborative practice communities were better able to recognise and adapt AIED's instructional value. Students reported greater engagement when AI functioned as a tutor, study companion, or Socratic opponent rather than merely a design aid. AI affordances were perceived as either challenging or encouraging depending on activity design. Based on these insights, the study proposes design dimensions for teachers to support responsible and pedagogically aligned AI integration into learning tasks.\n\n"]