This paper introduces a phase-geometric framework for understanding mental pathology as position, motion, and recoverability within an explicitly interpretable clinical phase space. Mental disorders are modeled not as categorical entities or symptom clusters, but as trajectories within a three-dimensional geometry defined by Structural Coupling (X), Energetic Drive (Y), and Semantic Proliferation (Z). Health is conceptualized as dynamic motion within a bounded region of recoverability—the Clinically Relevant Goldilocks Zone (CRGZ)—rather than as a single equilibrium point. To bridge subjective regulation and clinical structure, the framework establishes a non-invertible, many-to-one projection from high-dimensional IPCS micro-dynamics (Insight–Prediction–Control–Sociality) to macroscopic clinical geometry. A fourth dimension, Φ_Dark, captures chronicity and historical load as a modulation of navigability rather than an additional functional axis. Through comparative case geometries of obsessive–compulsive disorder, depression, dissociation, and schizophrenia, the paper demonstrates how distinct pathological regimes emerge as confined oscillation, contractive damping, fragmentation, and boundary instability within the same space. Recovery is formalized as ordered re-entry into the CRGZ, while pathology transitions are described as predictable trajectory redirections. The framework further clarifies the conceptual asymmetry between human hallucination and AI generative error by exposing a dimensional difference in historical load and coordinate stability. A multi-tier validation strategy is proposed, spanning self-report mapping, network-based neurophysiology, and longitudinal trajectory tracking. This work is part of the IPCSALT–UPF research program, which develops a unified phase-field approach to consciousness, psychopathology, and human–AI systems.