Coaching practices are rapidly changing due to advances in pervasive computing: behavior of coachees can unobtrusively be monitored in real time, and coaching can be remote or even fully automated. Fully autonomous e-coaching systems hold promise for improving people’s self-management, but also raise questions about the importance of human involvement in the e-coaching process. This paper describes an empirical ‘Wizard of Oz’ study in which coachees (N=82) were coached to take the stairs by either another person (N=20) or the e-coaching system eMate. Crucially, some coachees were made to believe that they would receive one type of coaching (human or computerized), while in reality they received the other. Results show that the coaching was equally effective in all groups, but that people who believed to be coached by a human judged the coaching to be more influential. No difference was found between groups in how trustworthy coachees found their coaches.