Abstract Classic Recharge Oscillator (RO) studies typically rely on spatially averaged indices, which capture the temporal characteristics of ENSO mode but provide limited information on its spatial structure. To address this limitation, a Recharge Oscillator‐based Linear Inverse Model (RO–LIM) was developed by embedding the recharge–discharge physics of ENSO into a data‐driven framework, enabling simultaneous diagnosis of its linear stability and spatial structure. Applied to reanalysis data, RO–LIM identifies a damped ENSO mode (decay rate of 0.29 years−1) for 1980–2023. After 2000, ENSO became more strongly damped (0.4 years−1) and transitioned from an eastward‐propagating eastern‐Pacific type to a westward‐propagating central‐Pacific or mixed type. This regime shift arises from weakened thermocline‐wind coupling but strengthened zonal current‐wind coupling under a La Niña‐like mean‐state change with enhanced trades and a steeper thermocline. RO–LIM provides a physically grounded and quantitative tool for diagnosing ENSO stability, spatial diversity, and their modulation by slow mean‐state change.