In the last decade, the educational field, in general, has experienced increasing interest in applying augmented reality (AR) for educational purposes. Studies have shown that when AR is effectively applied in education, it can increase students’ learning interest and concentration (Zhang et al., 2014), reduce cognitive overload (Bower et al., 2014, p.1), and provide a more authentic learning experience (Klopfer, 2008). This study uses both cognitive and sociocultural theoretical perspectives to better understand the role of AR in peer interaction by investigating language-related episodes (LREs) during collaborative dialogue. The current study investigates whether mobile-based AR influence the number, nature, outcome, and correction orientation of LREs during two oral and writing-focused activities of ten advanced L2 Spanish dyads using AR and non-AR mobile applications. The results show significant differences in the incidence of LREs in both settings (AR vs non-AR) and modality focus (oral vs writing-focused). Although significant differences were found between mechanical LREs vs. lexical and grammatical LREs, no significant differences were found between lexical and grammatical LREs in both modalities and settings. Likewise, the correction orientation was similar in both modalities, whereas the LRE outcomes were significantly different in both settings. Immediate posttests were administered to determine whether participants retained the results of the LREs based on the LRE outcome types. The posttests showed a strong correlation between the recognition and production scores of the grammatical structures. However, no significant differences were found in the recognition or production of grammatical structures nor the production of lexical items between the two settings.