SummaryWhen a high spectral efficiency is needed, the cost of Euclidean distance‐based antenna selection for spatial modulation (EDAS‐SM) in terms of hardware, size, and computational complexity is significantly increased because of the large transmit antenna array required. In comparison, generalized spatial modulation (GSM) can match the spectral efficiency of EDAS‐SM, while using significantly fewer transmit antenna elements. However, the error performance of GSM is naturally limited because of the use of a predetermined and fixed set of transmit antenna combinations. By exploiting knowledge of the channel, the optimal set of transmit antenna combinations can be selected by maximizing the minimum Euclidean distance between transmit vectors. In this paper, an adaptive scheme for selection of the optimal set of transmit antenna combinations is proposed to improve the reliability of GSM. The computational overhead of the said scheme is relatively high; hence, a low‐complexity suboptimal scheme for selection of the set of transmit antenna combinations is further proposed. The improved GSM schemes address the spectral efficiency limitation of EDAS‐SM, while demonstrating superior error performance.