In this study, the authors propose a low complexity user scheduling for multi‐antenna broadcast systems with a large number of users with diverse delay‐quality of service (QoS) assurances. Owing to the exclusive user scheduling constraints, the optimisation is combinatorial. Furthermore, by adopting effective capacity and effective bandwidth to illustrate the behaviours of different traffic characteristics such as different source statistics and queue dynamics, the delay‐bound violation probability constraints can be converted into equivalent minimum data rate constraints. To reduce the computational complexity, they use genetic algorithm (GA) to perform scheduling, instead of a brute‐force exhaustive search (ES) over all possible user subsets. By comparing the complexity of GA and ES, they show that GA is a rapid, although suboptimal, option of performing user scheduling optimisation. Simulation results show that the proposed algorithm can not only maximise the achievable user sum data rate, but also keep the delay‐bound violation probability of each user below a given threshold.