Using workplaces as a unit for randomization, a randomized controlled trial was conducted to determine the effects of web-based supervisor training on the subject of worksite mental health on job stressors, supervisor support and psychological distress among subordinate workers. A total of eight workplaces of a sales and service company were randomly assigned to either training workplaces or non-training workplaces. Supervisors (n=23) at the training workplaces participated in web-based self-learning training on worksite mental health; supervisors (n=23) at the non-training workplaces did not. A total of 81 subordinate workers under the trained supervisors (the intervention group) and 108 subordinate workers under the non-trained supervisors (the control group) completed the Brief Job Stress Questionnaire (BJSQ) at baseline and at a three-month follow-up. No significant intervention effect was observed for any scale of the BJSQ measuring job stressors, supervisor or coworker support, or psychological distress among subordinate workers (p<0.05). The item score of work autonomy changed very little in the intervention group, while it decreased in the control group during the follow-up period, yielding a significant intervention effect (p=0.02). The item score for a friendly atmosphere in the workplace increased in the intervention group, while the score remained stable in the control group, yielding a significant intervention effect (p=0.02). While the present study failed to show any clear effect of the web-based training of supervisors on reduction of job stressors, it may be useful for maintaining worker autonomy and improving the friendliness of the worksite atmosphere.