Visuospatial perspective-taking (VPT) is a fundamental process of social cognition. To date, few studies have investigated whether humans also take the perspective of humanoid robots. Recent findings are conflicting as one study found no evidence for level 1 VPT (i.e., which object is seen by the robot) and a different study found evidence for level 2 VPT (i.e., how the object is seen by the robot). The latter study proposed that the mere human-like appearance of robots triggers VPT (mere-appearance hypothesis). Here, we tested level 1 VPT for humanoid robots and the robustness of the mere-appearance hypothesis. We manipulated the human-like appearance of the robot and the robot’s mental capacity to perceive its environment. We found that all manipulations triggered VPT, showing, in contrast to earlier findings, level 1 VPT for robots. Also, our findings support the mere-appearance hypothesis as VPT was triggered regardless of whether the robot had a mental capacity to perceive its environment or not. In four follow-up experiments, we assessed the boundary conditions of the mere-appearance hypothesis and found that VPT is triggered not only by human-like visual features but also by non-human-like features implying a human presence.