The effective implementation of Collaborative Inquiry-based Learning (CIL) activities hinges upon the learners receiving ample support and timely feedback. Without explicit guidance from educators within the interaction design, inquiry processes, and task steps, the efficacy of learning outcomes is challenging to ensure. Virtual teachers have the capacity to substitute for human educators, offering learners knowledge-based answers and personalized support throughout the collaborative inquiry process. The present study employs a two-factor experimental design to investigate the impact of the anthropomorphic design features of virtual teachers (Photo-realistic, PVT vs. Cartoon-style, CVT) and the structure of collaborative scripts (Low, LCS vs. High, HCS) on the academic performance, motivation, and perceived needs of secondary school students engaged in CIL activities. Sixty-two students from two classes in a secondary school participated in the CIL activity experiment. Findings indicated that: (1) PVT significantly boosted learners' academic performance, while CVT enhanced learners' motivation and addressed their perceived needs. (2) LCS effectively stimulated learners' motivation and addressed their perceived needs, whereas HCS improved learners' academic performance. (3) There was no significant interaction between virtual teachers (PVT/CVT) and collaboration scripts (HCS/LCS) concerning learners' academic performance, motivation, and perceived needs.