Video communication via error-prone networks suffers from visibility of data impairment. Lots of research effort has been made to investigate error concealment algorithms, which almost always assume that all errors have been detected successfully. However, this assumption is not always the fact. As widely used, syntax and semantics analysis (SSA), cannot guarantee complete error detection. Furthermore, an error detected by SSA should not be inevitably concealed unless it is a visual artifact. On the other hand, temporal concealment, using motion vector recovery (MVR) and copying, has been recognized simple and effective. This paper proposes an MVR based error and artifact detection (EAD) algorithm, abbreviated by MVR-EAD, which compares a risk area with its temporal counterpart in the reference frame and decides if an artifact occurs. MVR-EAD gains attractive detection accuracy, an increase from 7.8 to 73.87 % demonstrated by extensive experiments. Accordingly, this incurs 0.09 to 1.88 dB in PSNR improvement with the error concealment strategy embedded within JM18.0.