Background: A growing body of anecdotal evidence indicates that the use of robots may provide unique opportunities for assisting children with autism spectrum disorders (ASD). However, previous studies investigating the effects of interventions using robots on joint attention (JA) in children with ASD have shown insufficient results. The robots used in these studies could not turn their eyes, which was a limitation preventing the robot from resembling a human agent. Methods: We compared the behavior of children with ASD with that of children with typical development (TD) during a JA elicitation task while the children interacted with either a human or a robotic agent. We used the robot "CommU," which has clear eyes and can turn its eyes, for the robotic intervention. The age range of the participants was limited to 5-6 years. Results: Sixty-eight participants participated in this study, including 30 (10 females and 20 males) children with ASD and 38 (13 females and 25 males) children with TD. The participants were randomly assigned to one of the following two groups: the robotic intervention group or the control group. JA in the children with ASD was better during the robotic intervention than during the human agent intervention. These children exhibited improved performance in the JA task with human after interacting with the robot CommU. JA was differentially facilitated by the human and robotic agents between the ASD and TD children. Conclusions: The findings of this study significantly contribute to the literature on the impact of robots on JA and provide information regarding the suitability of specific robot types for therapeutic use.