Automated Fiber Placement (AFP) of carbon-fiber composite materials constitutes a core digital manufacturing technology for large-scale aerospace composite components with double-curved surfaces. Gap and overlap defects induced during path planning degrade the fatigue performance of composite parts and threaten aircraft structural safety. Aiming at the limitation that conventional post-forming non-destructive testing cannot predict lay-up defects at the design stage, this work analyzes the generation mechanisms of gaps and overlaps from differential-geometric constraints of non-developable surfaces. A defect classification framework and geometric identification method oriented to the path-planning phase are established. According to geometric formation causes, lay-up defects are categorized into boundary overlap–gaps induced by tow-cutting operations and internal gaps originating from trajectory divergence over curved surfaces. Twenty boundary working conditions are defined for triangular boundary defects, and internal gaps are further subdivided into tape-type gaps and rectangular gaps. A triangular-patch approximation algorithm is adopted to compute the area of gap–overlap regions and quantitatively evaluate lay-up coverage. The simulation results reveal an inherent trade-off between allowable fiber-angle deviation and lay-up coverage on double-curved mold surfaces.