The reconstruction process of structured illumination microscopy (SIM) creates substantial artefacts if the specimen has moved during the acquisition. This reduces the applicability of SIM for live cell imaging, because these artefacts cannot always be recognized as such in the final image. A movement is not necessarily visible in the raw data, due to the varying excitation patterns and the photon noise. We present a method to detect motion by extracting and comparing two independent 3D wide-field images out of the standard SIM raw data without needing additional images. Their difference reveals moving objects overlaid with noise, which are distinguished by a probability theory-based analysis. Our algorithm tags motion-artefacts in the final high-resolution image for the first time, preventing the end-user from misinterpreting the data. We show and explain different types of artefacts and demonstrate our algorithm on a living cell.