Resilience and flow are crucial in language education, yet most research focuses on formal learning environments, with limited studies on their impact in informal settings. This study explores the relationship between basic psychological needs and engagement in the context of informal digital English learning (IDLE). Using a mixed-methods design, data were collected from 512 Chinese EFL learners. Structural equation modeling and NVivo analysis were applied to quantitative and qualitative data, respectively. The findings reveal that resilience fully mediates the relationship between basic psychological needs and engagement, serving as an adaptability enhancer, persistence promoter, stress buffer, self-efficacy builder, and emotional regulation facilitator. Conversely, flow partially mediates this relationship, though perceived competence does not significantly predict flow in this context. Building on this, flow contributes as an intrinsic motivation driver, positive cycle creator, external pressure mitigator, and efficiency enhancer. This research underscores the important role of resilience and flow in IDLE among Chinese university EFL students. By highlighting these mediating roles, the study provides valuable insights for enhancing the effectiveness of IDLE experiences, contributing to the broader discourse on language education in the digital age.

The rapid advancement of artificial intelligence (AI) has transformed informal language learning, providing learners with flexible, self-directed opportunities for vocabulary acquisition outside formal classroom instruction. This study examined how AI use, teacher support, and periodic testing influence Chinese EFL learners’ engagement in informal English vocabulary learning. A sequential mixed-methods design was employed, combining a three-group pretest-posttest quantitative intervention over three months with follow-up qualitative interviews involving 15 participants. Quantitative results showed significant increases in all three engagement dimensions across conditions. At posttest, the teacher-guided AI with testing condition produced the highest behavioral and cognitive engagement, whereas the teacher-guided AI condition produced the highest emotional engagement; the AI-only condition showed the lowest levels across all dimensions. Qualitative accounts suggested that teacher support facilitated goal-setting, strategic scaffolding, and emotional reassurance, helping learners use AI more purposefully, while participants described testing as providing structured feedback for strategic adjustment and self-monitoring. The AI tool supported deep lexical processing and personalized learning, while learners appeared to use these affordances more strategically when combined with teacher guidance and structured feedback. These findings are broadly consistent with Self-Determination Theory and provide practical guidance for designing learner-centered vocabulary programs that balance autonomy, guidance, and accountability while sustaining strategic and emotional engagement.