The impact of Pi application as an AI-assisted speaking assessment tool on EFL learners’ overall speaking perceptions, foreign language anxiety, speaking self-efficacy, and speaking satisfaction

This study examined whether Pi Application (PiApp), an AI-assisted speaking assessment and practice tool, was associated with short-term changes in EFL learners’ overall speaking perceptions (OSP), foreign language anxiety (FLA), speaking self-efficacy (SSE), and speaking satisfaction (SS). Using a randomized pretest–posttest control-group design, 30 intermediate EFL learners in Iran were assigned to either a PiApp-supported experimental group or a conventional classroom control group (n = 15 each). During the eight-week intervention, both groups completed comparable speaking tasks, but the experimental group received PiApp-generated formative feedback, whereas the control group received teacher-mediated feedback. Data were collected through Likert-type scales and analyzed using MANCOVA, with pretest scores as covariates. A significant multivariate group difference was found, Pillai’s Trace = 0.889, p < .001. Follow-up ANCOVAs showed that, after baseline adjustment, the PiApp group reported lower FLA and higher OSP, SSE, and SS than the control group. These findings provide preliminary evidence that PiApp-supported speaking practice may improve learners’ self-reported affective, perceptual, and self-belief outcomes. However, generalizability is limited by the small sample, short intervention, absence of delayed follow-up, and lack of direct speaking-performance assessment.