In this paper, we propose a robust transceiver design for nonregenerative multicasting multiple-input multiple-output (MIMO) relay systems where a transmitter broadcasts common message to multiple receivers with aid of a relay node and the transmitter, relay and receivers are all equipped with multiple antennas. In the proposed design, the actual channel state information (CSI) is assumed as a Gaussian random matrix with the estimated CSI as the mean value, and the channel estimation errors are derived from the well-known Kronecker model. In the proposed design scheme, the transmitter and relay precoding matrices are jointly optimized to minimize the maximal mean squared-error (MSE) of the estimated signal at all receivers. The optimization problem is highly nonconvex in nature. Hence, we propose a low complexity solution by exploiting the optimal structure of the relay precoding matrix. Numerical simulations demonstrate the improved robustness of the proposed transceiver design algorithm against the CSI mismatch.