An IR thermography-based detection and localization of turbulent flow separation at an operating wind turbine is presented and verified for the first time. Turbulent flow separation limits the efficiency of wind turbines and causes increased structural loads and acoustic emissions. IR thermography is an established measurement method for stall detection in wind tunnel experiments, however a transfer to operating wind turbines is an open research question. With respect to the state of the art for thermographic stall detection, a novel thermographic measurement approach for a feature-based stall detection is presented, verified and applied on wind turbines. The measurement approach evaluates the surface temperature response to unsteady inflow conditions and enables an unambiguous detection of flow separation by means of temperature fluctuation maxima in the regions of flow transition as well as an increasing temperature fluctuation within the separated flow region. Finally, the aim to obtain a non-invasive and in-process detection of flow separation on an operating wind turbine with IR thermography is achieved and verified using tufts flow visualization.