<p>While preparing mathematics learning activities is one of the most challenging tasks for teachers, creating outdoor learning activities, in particular, poses additional difficulties due to a lack of expertise and resources. Artificial intelligence can facilitate the planning processes of future educators, thereby enabling them to deliver outdoor learning activities that are well-organized and align with the curriculum. This study investigated pre-service mathematics teachers’ use of ChatGPT during the design of outdoor learning activities. It specifically aimed to analyze the properties of the prompts they used, including prompt types and content, and to determine the relationship between prompt quality and the quality of the resulting AI-supported activities. A qualitative case study design was conducted with 34 pre-service mathematics teachers. The quality of the AI-supported activities was evaluated using a rubric with three dimensions: environment, learner, and educator. The findings show that several groups produced medium- and high-quality activities during an AI-supported design process. The use of two-dimensional and multi-dimensional prompts was associated with higher activity quality, whereas prompt type (single vs. multiple) was not directly related to quality level. Based on these findings, teacher education programs may benefit from incorporating training on prompt-related competencies, including prompt design and the critical evaluation of AI-generated outputs.</p>