How L2 Learners Negotiate Meaning in GenAI-Supported Creative Writing

This qualitative study explores how second language (L2) learners negotiate meaning and co-construct knowledge with generative AI (GenAI) in a 12-week multimodal creative writing project. Chinese middle school students (N = 75) created English picture books using a conversational GenAI agent supporting textual and text-to-image generation. By extreme-case sampling, 28 focal participants (14 high- and 14 low-performing) were selected based on creative product quality. Student-AI textual interaction logs were analyzed using the Negotiation for Meaning (NfM) framework and Gunawardena and colleagues’ Interaction Analysis Model (IAM), supplemented by artifact analysis and interviews. Results revealed distinct patterns: high-performing learners showed active, evaluative engagement with longer prompts, higher NfM completion rates, and deeper IAM phases. They treated GenAI as a scaffold for linguistic and multimodal refinement while retaining strong authorial control. Low-performing learners, conversely, displayed passive patterns with vague prompts and greater AI content retention, yielding a weaker personal voice. Findings suggest creative outcomes in GenAI-assisted L2 multimodal writing depend more on depth of meaning negotiation and strategic mediation than interaction frequency. The study extends interactionist and sociocultural perspectives to GenAI-mediated creative genres and underscores scaffolds for negotiation competence and authorial agency.