The Effects of Generative AI, Peer Interaction, and Interaction Structure on Oral Performance and Perceived Feedback Among Beginning Learners of Chinese as a Foreign Language

ABSTRACT While the application of generative artificial intelligence (GAI) in language learning is rapidly expanding, its comparative effectiveness with peer interaction for developing oral performance among beginner learners of Chinese as a foreign language (CFL) remains underexplored, particularly regarding the critical variable of interaction structure. This study adopted a 16‐week randomized explanatory mixed‐methods design to systematically examine the effects of interaction partner (GAI vs. peer) and interaction structure (strong vs. weak) on CFL learners’ oral performance and perceived feedback. Results indicate that interaction structure, rather than interaction partner, is the primary factor influencing oral production. Strong interaction structure significantly improved fluency, accuracy, discourse organization, topical development, perceived usefulness, and communicative growth, and further activated self‐monitoring and feedback mechanisms under the GAI condition. GAI interaction enhanced fluency and emotional safety, while peer interaction fostered discourse coherence and communicative naturalness. Interview data supported the complementary roles of GAI and peer interaction. The study recommends optimizing interaction structure and task design to support the development of a beginner‐level oral instruction model that integrates timely, individualized feedback with collaborative learning.