Answers are grounded in this paper and cite their sources.
AI-mediated speaking practice may provide possibilities for practice with less interpersonal pressure, yet speaking anxiety is still a persistent challenge in EFL learning. This quasi-experimental mixed-methods study examined whether additional AI-mediated speaking practice was associated with changes in speaking anxiety among Indonesian EFL students and explored students’ affective experiences during such practice. Forty-five students from two intact classes participated: 25 completed 24 additional AI-mediated speaking sessions over eight weeks, while 20 received regular speaking instruction. Speaking anxiety was measured before and after the intervention, with posttest scores analysed using ANCOVA controlling for pretest anxiety. Twelve experimental-group students were subsequently interviewed. Both groups showed small descriptive decreases in anxiety, but ANCOVA detected no between-group difference at posttest after controlling for baseline anxiety (p = .977, ?p² < .001). The qualitative data indicated three patterns of affective experience: (1) students reported lower social-evaluative pressure during AI-mediated speaking practice; (2) feedback became consequential to students due to learner uptake and repeated practice; and (3) although students reported increased confidence in their speaking abilities, this was not generalised over contexts. These experiences varied across participants and were affected by students’ remaining learning anxieties, variation in uptake of the AI, and varying degrees of technological frictions. Integrating the findings suggested complementarity rather than contradiction: positive affective experiences were reported during AI-mediated speaking practice, but between-group difference in speaking anxiety was not detectable. AI may offer a relatively low-pressure environment for speaking practice, but context specific comfort and confidence should not be taken as evidence of broader anxiety reduction.
No takes yet. Share an insight, caveat, or question.
Budiawan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: