The act of reflecting on physically active possible selves can increase physical activity (PA). According to theory, possible selves that include strategies for achieving them (self-regulatory possible selves) should have the greatest impact on behavior. Our aim was to examine whether forming a self-regulatory physically active possible self is more effective at increasing PA than forming a possible self that focuses only on the image of the possible self (self-enhancing possible self) or engaging in a control activity. Task self-efficacy was examined as a moderator.An online, randomized experimental study.Insufficiently active adults (n = 244) completed task self-efficacy and PA measures before the intervention, and 4 and 8 weeks after.Repeated measures ANOVA revealed a 'time by condition' interaction whereby reported PA levels were higher for participants in the self-enhancing condition than for those in the control condition 4 weeks post-intervention. Reported PA levels were also higher for participants in the self-regulatory condition than those in the control condition at both follow-up time points. There was also a 'time by condition by self-efficacy' interaction, whereby participants in the self-enhancing condition reported more PA than controls at both follow-up points when they also reported high self-efficacy. Participants in the self-regulatory condition reported more physical activity than all other participants when they were also low on self-efficacy, but only at the four-week follow-up point.The findings extend the PA possible selves literature by suggesting that different types of possible selves interventions may work best depending on participants' task self-efficacy levels.