Artificial agents, such as voice-controlled conversational agents (VCAs) built into smart devices, are becoming more prevalent in daily and educational contexts, enhancing the possibility of using them as language partners. However, research has primarily focused on the cognitive or affective outcomes of using these agents, overlooking questions about language learners' perceptions of agents as human-like conversational partners and about what aspects of agents generate the social interaction schema required for language learning. To address these gaps, this exploratory study examined to what extent EFL students perceived a VCA as a human-like language partner and what knowledge regarding language partners these students used to justify their perceptions. Sixty-seven Korean EFL students, all of them being nine years old, completed three interactional tasks with a VCA designed to act as a language partner. They then participated in a drawing task and in-depth interview implemented to explore students' perceptions toward the VCA as a language partner. Thematic analysis of students' drawings and interview transcripts found that the majority of students identified human elements in the VCA, perceiving it either as a human-like partner or as something between artifact and human. This strong tendency toward anthropomorphism indicates VCAs' great potential as interactive language partners in EFL contexts. Additionally, this study discusses the reasons for and implications of students' strong anthropomorphism tendency and how students used their knowledge regarding language partners to justify their VCA perceptions. Future recommendations for the use of VCAs were suggested from both pedagogical and technological standpoints.