Abstract: Faced with the complex urban diseases that emerge within territorial space, territorial spatial diagnosis often falls into fragmented symptomatic treatment and hurried experience-based decision-making during fast-moving planning practice, which may turn spatial planning into an abstract exercise detached from reality. Although AI technologies with higher-level holistic reasoning and ecological rationality show substantial potential for improving the scientific rationality of territorial spatial planning, effective theoretical paradigms and operational methods for AI-assisted territorial spatial diagnosis remain insufficient. Starting from the theoretical characteristics and core tasks of territorial spatial diagnosis, this paper reviews the development trajectory of diagnostic technologies and identifies the challenges facing their intelligent transformation. On this basis, it constructs a Dual-Helix paradigm and a Diamond framework for territorial spatial diagnosis under human-machine hybrid intelligence, and further proposes five directions for changes in cognitive and action methods and six systems for enhancing working methods after AI is introduced into territorial spatial diagnosis. The proposed framework is expected to systematically address the problems and symptoms of traditional territorial spatial diagnosis, improve the scientific rationality of territorial spatial planning, and promote the sustainable development of territorial space.