The humanoid robot TEO is an assistive household companion which is able to represent Spanish Sign Language (LSE) vocabulary and dactylology thanks to its underactuated robotic hands. Prior studies on sign language representation through the simulation of the humanoid robot TEO revealed that end-users’ predisposition to communicate with robots via sign language was over 80% positive. Moreover, around 65% of reticent end-users changed their minds after their first experience with the simulated robot. Unexpectedly, a new developed study has shown that embodiment has dropped satisfaction rates drastically and increased comprehension rates. The under-actuated movement of the real hands has been modelled according to three generators based on three different neural networks, and the data obtained in previous simulation. Therefore, four different dactylology models have been shown to end-users. It has to be considered that the notion of embodiment is required where there is cohesive interaction between the environment and the body. Despite this, recent end-user feedback has shown some recurring criticisms referring to the embodied robot that did not arise with the simulation experiments. Among the most recurrent topics, the demand of facial expression is emphasised. In terms of demographics, the decision of grouping the academic background groups into two main sectors (university and non-university studies) relies on the link between university and research. Therefore, university students and graduates are expected to be more aware of actual robotics state of the art and, consequently, to be more critical towards the scientific implementation. As expected, lower satisfaction rates are detected among university students and graduates. Differences in comprehension rates are not significant. Nevertheless, a more pronounced overall age-related comprehension decreasing trend is identified for non-university graduates.