This study aimed to examine an augmented reality-based teaching simulation in a mobile application. We examined how AR-enabled interactions affect users’ perceived scenario usefulness and avatar realism. The participants were forty-six undergraduate students. We randomly grouped them into two conditions: AR and Non-interactive video groups with equal sample sizes. This study employed an experimental design approach with a one-way multivariate analysis of variance with repeated measures. The independent variable is the presence/absence of AR interaction with a mobile application. The dependent variables were avatar realism and scenario usefulness. The measures explored how the student avatar’s emotional intensity in a scenario influences user perception. The results showed that participants in the AR-interaction group perceived avatar realism significantly higher than those in the non-interactive video group. Also, participants perceived the high emotional intensity scenario (aggression toward peers) to be significantly higher usefulness than the low emotional intensity scenario (classroom disruption).