A closed-loop AI virtual simulation platform for preclinical fiber post-core restoration: two-year prospective evaluation of learning outcomes and student perceptions

Preclinical prosthodontic education faces challenges in cultivating the integrated clinical decision-making and operational skills required for complex procedures like fiber post-core restoration. While virtual simulation (VS) is widely used, most systems function primarily as skill trainers and lack support for systematic clinical reasoning training. This two-year prospective study aimed to report implementation‑related observations and evaluate student acceptance of a novel AI-enhanced virtual simulation (AI-VS) platform integrating dynamic AI virtual patient consultation, domain‑specific procedural Q&A assistant, and a standardized rubric-based medical record evaluation workflow for comprehensive fiber post-core restoration training. This pre-post design included two independent fourth-year dental student cohorts: 2024 ( n = 38) and 2025 ( n = 42). The AI-VS platform provided a complete training cycle. Objective outcomes included pre-and post-training clinical case analysis scores (assessing clinical reasoning) and post-training simulator operation scores (assessing procedural performance at course completion). Subjective experience was measured using a study‑specific questionnaire developed for this investigation. Face‑and‑content validity assessment was performed via internal review by the three study authors who also developed the questionnaire; this represents internal‑team assessment and does not constitute independent external expert validation. Internal‑consistency reliability assessment was performed for multi‑item dimensions; survey constructs covered learning interest/satisfaction, self‑perceived outcomes, career expectation, and AI acceptance. Following curriculum implementation, both cohorts showed statistically significant improvements in clinical reasoning scores compared to baseline (2024: median 69 to 81.5; 2025: 71 to 83; both p < 0.001). A strong negative correlation was found between pre-training scores and score changes (2024: ρ = -0.879; 2025: ρ = -0.789; both p < 0.001). However, this pattern should be interpreted with caution, as it may reflect statistical artifacts including regression to the mean, ceiling effects, and the mathematical coupling inherent in correlating baseline scores with change scores. Post-training clinical reasoning and operational scores were strongly positively correlated (2024: r = 0.927; 2025: r = 0.722; overall: r = 0.840; all p < 0.001), suggesting a positive association between decision-making and procedural performance at course completion. Subjectively, students reported high course satisfaction (mean = 4.26/5), strong subjective endorsement of AI’s utility‑‑particularly its perceived ability to compensate for limited one‑on‑one teacher guidance (mean = 4.29/5)—and very positive career expectations (mean = 4.34/5). Following implementation of the AI‑VS curriculum, clinical‑case analysis scores increased in both student cohorts, post‑training procedural performance was positively associated with clinical‑reasoning scores, and students reported high acceptance of the course. Because the study used an uncontrolled pre‑post design and procedural performance was assessed only after training, these findings should be interpreted as preliminary implementation‑related observations rather than evidence of platform effectiveness, educational equity, or improvement in procedural skills.