<p>While AI literacy is commonly framed through competency-based or ethical frameworks, far less attention has been given to AI literacy as a design-oriented epistemic process. Drawing on Learning by Design, and Design based research, this study treats AI-mediated activity as a site of meaning-making, reflexive judgement, and pedagogical agency. As teacher agency is enacted through everyday design decisions, I explore how GenAI mediates lesson design in science teacher education and how such mediation reshapes emerging forms of self-regulation and assessment-as-learning. Informed by the pedagogy of multiliteracies and Learning by Design, the study brings together two interconnected empirical strands: close analysis of prompt-based interactions during AI-supported lesson redesign and reflective texts produced alongside these design activities. Read together, these materials show that engagement with AI is not a peripheral aid but a constitutive part of pedagogical reasoning in lesson design. Lesson redesign through cycles of proposal, interrogation, revision, and selective rejection of AI-generated suggestions, activated forms of planning, monitoring, and evaluation typically associated with self-regulated learning and assessment-as-learning, while situating them within concrete design decisions. Reflection did not serve as an after-the-fact commentary on completed work but it was embedded in the act of designing itself. With Human-AI interaction as a design space in teacher education, enacted AI literacy reframes assessment as epistemic visibility: a way of making teachers’ reasoning, judgement, and responsibility observable; thus contributing to current debates on assessment education and showing how AI-mediated design can support process-oriented, ethically grounded evaluation of learning in science teacher education. </p>