The characteristics of human–human and human–animal interactions have been studied intensively. Previous studies showed that the presence of a third human party can influence motivation during a cognitive task, whereas a third animal party, not expected to make any judgement, does not have a similar impact. As a growing number of animated objects are created to interact with humans, we asked whether humans can perceive robots as potential interlocutors. The question of how these objects should be visually designed in order to be accepted by human users as interlocutors and assistants remains open, essentially owing to a lack of behavioural surveys. Therefore, the present study compared the impact of the presence of different animated objects on the behaviour and performance of primary school children during a cognitive task. Fifty-one primary school pupils (10 years 11 months ± 6 months) were individually given mental arithmetic tasks ‘supervised’ by a computer alone, or by the same computer assisted by an animated object (i.e. virtual human character, animal-shaped object covered with fur, humanoid metallic robot). Children's performances and behaviour were analysed. First, we found that animated objects were accepted as interacting partners, as they modified children's attention and influenced their emotional state in a positive way. Second, we found that the object design (human-, animal- or robot-like) influenced the children's reactions and performance in different ways. 3D objects (animal and robot), with more ‘live’ characteristics (e.g. motion), elicited more positive behaviours as well as fewer negative behaviours than the 2D character. However, the animal-like object decreased attention towards the computer more, and consequently also cognitive performance, than the human or humanoid ones. This study opens new research perspectives concerning the importance of the realism of robotic objects and the associated capacity of judgement on the level of acceptance by humans.