Trust plays a crucial role in Human-Robot interactions. However, the prevailing belief that robots are inherently capable can lead to overtrust, where individuals rely on robots’ decisions without applying critical thinking. To evaluate the extent of overtrust in HRI, we conducted an experiment in which participants engaged in a cognitive task alongside a simple robotic object that performed the same task. Participants’ performances were compared to a baseline group in which the robot moved but didn’t provide answers to the task items. We deliberately chose a simple non-humanoid robot and informed participants that the robot was in a training phase where it was learning how to perform the task. The results showed that when the robot provided answers, participants tended to align with them, even when they were wrong. Additionally, participants’ reaction time was shorter, indicating they were investing less effort in solving the task. Participants also demonstrated decreased self-efficacy and were less confident in their performances. Our findings indicate that overtrust should be carefully considered in the design and implementation of even very simple robots.