Driven by the wave of artificial intelligence, the educational practice and application of robots have become increasingly common. Despite extensive coverage in the literature on various aspects of educational robots, there are still unexplored avenues, particularly regarding robotic support, robotic personality, and challenges in their applications. This study presented a systematic review of high-quality empirical research on the use of physical robots in educational settings. A total of 92 relevant papers from the Web of Science database were analyzed. Employing the technological pedagogical content knowledge (TPCK) framework, we investigated research questions across seven components, including the learning domain, teaching strategy, robot types, learning results, problems with using robots, robotic support, and robotic personality. The findings revealed that robots are most prevalently employed in language learning applications. When opting for teaching strategies, educators tend to favor those that incorporate physical interaction. Concurrently, humanoid robots emerge as the preferred choice among many. These robots, in human–robot interaction scenarios, often exhibit an agreeable personality. In terms of evaluating learning results, cognitive aspects like thinking, creativity, self-regulation, and inquiry ability are especially emphasized. Such results are frequently influenced by the informational and emotional support provided by robots. Nonetheless, challenges encountered by teachers, learners, and robots in this process are not to be overlooked. The findings of this study contributed to future applications of robotics in education.

The Chinese language has unique tonal phonology and script characteristics that pose specific problems for international students learning Mandarin, and the ways in which human and AI support can complement each other to activate students’ learning motivation are not well understood. Building on control-value theory (CVT), this study constructs models of the effects of Chinese-language instructor support on Mandarin learning motivation and of the mediating role in the process, which are then verified and strengthened through chatbot immersion in Mandarin. Based on CVT, this study constructs models of the effect of Chinese-language instructor support on Mandarin learning motivation and the role of the mediator in this process, and then verifies and strengthens these models through chatbot immersion in Mandarin. Survey responses were obtained from 406 degree-seeking students in four regional university hubs (Xi’an, Wuhan, Nanjing, and Hangzhou) and analyzed using PLS-SEM. The measurement model revealed good convergent, discriminant, and reliability. Instructors’ support had a direct effect on motivation (β = 0.263) and an indirect effect through learning buoyancy (β = 0.043) and self-concept (β = 0.108), both mediations being significant. Each indirect path had a positive moderating effect of chatbot immersion, which strengthened the paths. The full model accounted for 34% of the variance in learning motivation. Results not only extend CVT to a logographic SLA setting but also demonstrate the complementary nature of human scaffolding and AI practice, providing practical recommendations for teacher training, resilience training, and technology integration in Chinese universities.