The rapid development of AIGC makes it very important to study efficient AIGC video detection methods. Compared with the detection methods that detect low-level artifacts and frequency-domain features, the detection methods that utilize high-level semantic features have better detection and generalization effects. This study finds that in some AIGC videos, the situation of missing motion details occurs. Therefore, this study proposes a method for AIGC video detection using the absence of motion details. This method extracts the motion features of animals and humans through 3D keypoint detection, and determines whether there is the absence of motion details by calculating the cosine similarity of the relative position vectors of adjacent positions in two adjacent frames. In the experiments, an AIGC video dataset of missing motion details is constructed, based on which the proposed method outperforms the state-of-the-arts.