In contrast to qualitative MRI, motion artifacts can be more subtle in quantitative MRI methods such as Magnetic Resonance Fingerprinting (MRF). Errors caused by motion are not easily detectable by visual inspection of resulting maps. Hence, there is clear need for supporting the reliability of results with regard to motion-induced errors. We present a method to detect if significant through-plane motion occurred during an MRF scan, without external motion tracking devices or acquiring additional data. The method is based on classifying the spatiotemporal residuals either by eye or a neural network. The performance was successfully evaluated in a patient study.