Distributed generation (DG) has reformed the meaning of power generation from large scale to small scale, but unintentional islanding is the main issue when connecting DG and the utility grid. A lot of techniques have been used for detecting islanding, among these techniques, there are passive and active. The main problem of passive techniques is their large non-detection zone (NDZ), while the main drawback of active techniques is their undesirable effect on power quality. In this paper, a proposed hybrid passive–active systematic methodology based on a smart classifier that decides to use an active method instead of a passive one is presented. In the proposed scheme, sensors are used for measuring the reactive power at three terminals: the DG terminal, grid terminal, and load terminal. The novelty in this paper is the accurate detection of islanding within a shorter time either in the normal case or NDZ; also it can differentiate between islanding and grid faults without degrading the power quality of the overall system as the active technique does not have to be used continuously, and so total harmonic distortion does not exceed the standard value (5%) detected by IEEE standards. The proposed scheme was simulated using the MATLAB/Simulink platform, and the results reflect its potential with a comparative study.