Recent studies have explored the determinants of acceptance of interactive dialogue systems such as robots and virtual agents, highlighting the importance of their physical and social characteristics and their use in different social contexts. This preliminary investigation aims to understand children’s preferences towards different types of virtual agents and a humanoid robot through the identification of tasks for which they are considered best suited for. It will also investigate whether and how participants’ s gender and agents’ features can influence these possible preferences. The project involved 20 children aged between 7 and 9 years. The experiment consisted of showing 13 video clips lasting 4 s. These video clips were purposely created and depicted 4 human agents (2 adults and 2 children, gender-balanced, with human and synthetic voices), 2 types of animal agents (dog and cat respectively, each species has been gender-balanced, with human voices), and a humanoid robot video clip Pepper. Preferences were assessed using a reduced and modified version of the Virtual Agent Voice Acceptance Questionnaire (VAVAQ), a new and original version compared to the previous ones, specifically designed for children users. Outcomes highlight a strong preference for animal agents for play activities, whereas human agents were considered more qualified for educational tasks.