This study examines the difficulties faced by content experts in the ontology development process for science education and explores issues encountered in collaboration with developers. The findings aim to provide foundational data for AI-based learning platform development. Using qualitative analysis (Lecompte & Preissle, 1993), the study investigates the experiences of ten content experts involved in constructing a middle school science ontology. The difficulties encountered by content experts were categorized into three areas: entity setting, relation setting, and attribute setting. In entity setting, difficulties arose due to ambiguous curriculum descriptions, differences in curriculum interpretation among experts, selection of entities related to scientific inquiry, and lack of interdisciplinary understanding. Relation setting required discussions on entity dependency, parallelism, and alignment with the learning sequence. In attribute setting, issues included polysemous terminology and defining the depth and scope of learning. Similarly, collaboration with developers presented difficulties in these areas. Entity setting issues included curriculum structure by grade level, duplicate terminology, and errors from special characters. Relationship setting revealed inconsistencies between the curriculum’s formal hierarchy and actual learning sequences, along with differences in conceptual understanding. Attribute setting faced visualization constraints due to attribute length limitations. These findings provide insights for the systematic development of AI-driven personalized learning platforms, emphasizing the need for enhanced interdisciplinary collaboration.