The control of human movements is thought to automize with repetition, promoting consistent execution and reduced dual-task costs. However, contingencies such as illness or constraints to regular movement patterns can promote conscious motor control, which can reduce movement proficiency and make dual-task situations more difficult. This experiment evaluated whether electroencephalographic neurofeedback training can reduce the adverse effects of conscious motor control. Twenty-five participants completed the timed-up-and-go task while wearing a leg brace to de-automize their regular movement, under both single and dual-task (walking + serial sevens) conditions, both before and after 30-min of neurofeedback training. Three different types of neurofeedback were prescribed across three laboratory visits. We hypothesised that training to decrease central EEG alpha-power at scalp sites above the supplementary motor area would facilitate performance compared to opposite (increase central EEG alpha-power) or sham neurofeedback training. Results revealed a pre-test to post-test improvement in performance on the single-task and on both aspects of the dual-task when participants were trained to decrease central EEG alpha-power. There were no benefits of opposite or sham neurofeedback training. Mediation analyses revealed that the improvement in dual-task motor performance was mediated by the improvement in cognitive performance. This suggests that the neurofeedback protocol was beneficial because it helped to reduce conscious control of the motor task. The findings could have important implications for rehabilitation and high-performance (e.g., elite sport) domains; neurofeedback could be prescribed to help alleviate the problems that can arise when individuals exert conscious motor control. Electronic supplementary material The online version of this article (10.1007/s00221-020-05935-3) contains supplementary material, which is available to authorized users.