In this study, the Bystander Turing Test (BTT) paradigm is used to determine whether participants rate particular dialog moves in tutoring transcripts to be generated by a computer tutor, AutoTutor, or by a skilled human tutor. Currently, AutoTutor initiates a conversation with the learner by posing a question or presenting a problem for the learner to solve. Students learn about computer literacy or physics by engaging in a conversation with an animated agent that remains on the monitor throughout the tutoring session. AutuTutor scaffolds the conversation with a series of dialog moves that are frequently used by effective human tutors. Results indicate that participants are unable to differentiate computer-generated moves from those generated by skilled human tutors. Participants did, however, view some dialog move categories to be more pedagogically effective than others regardless of whether the dialog move was generated by AutoTutor or by a skilled human tutor. (Contains 17 references.) (Author)