This paper focuses on a method of attention bias training, considering in particular its efficacy and usability in several mental disorders. The results of current meta-analyses and selected experiments indicate possible efficacy of training in case of some anxiety disorders (generalised anxiety disorder and social phobia), particularly in young individuals. Its efficacy in other previously tested disorders such as depression and addictions seems questionable. We analysed moderators of training efficacy considered in previous studies: subjects’ age, type of training task, type and location of emotional stimuli, duration of training, awareness of test objective and place of testing (research laboratory or subjects’ homes). It seems that greater efficacy of attention bias can be achieved by conducting longer trainings, located in a laboratory, rather than in-house, and using verbal rather than visual stimuli. It is not clear whether participants should be informed of the training objective or whether arranging stimuli vertically is more efficient than horizontally.