<p>The rapid development of generative artificial intelligence technologies presents new opportunities and challenges for learners, teachers, and pre-service teachers in education. While previous studies have focused on generative artificial intelligence (GenAI) in education, including teachers’ and learners’ experiences, less attention has been paid to how pre-service English language teachers (PELTs) incorporate GenAI into their materials design. Drawing on an ecological perspective on teacher agency, this qualitative study explores four PELTs’ experiences in designing GenAI-based teaching materials for their future classrooms. We collected multiple forms of qualitative data, including reflective journals, semi-structured interviews, and photovoice artifacts, which were thematically analyzed. Findings are Findings are organized into two main areas: (a) pre-service teachers’ negotiation of the affordances and constraints of generative artificial intelligence in developing instructional materials and (b) their agentive practices in adapting materials, designing lessons, and developing assessment practices. Across cases, prior learning histories, institutional conditions, and future aspirations shaped how PELTs appropriated GenAI to support autonomy, multimodality, and formative assessment, while also grappling with plagiarism, bias, privacy, authorship evidence, and uneven digital infrastructures. The findings suggest that preparing PELTs for safe and effective use of GenAI requires explicit ethics policies, assessment designs that make learner authorship visible, and collaborative structures that build critical artificial intelligence literacy. This study contributes a practice-anchored account of GenAI-mediated ELT and shows how teacher agency can illuminate the pedagogical and ethical negotiation involved in materials design.</p>