Interactional metadisocurse (MD) features in academic writing have gained much currency in today’s research. These features are manifested through various linguistic devices which express epistemic stance and sensitivity to audience. Although a large body of research on interactional MD has been produced, little attention was given to the study of these features in L2 writing in the Arabic context. This paper examines the extent to which Arab L2 writers deploy interactional MD strategies in academic writing. In this context, a corpus of 20 applied linguistics research articles (RAs) were compiled. Following Hyland’s MD taxonomy, interactional MD features were identified and analyzed. Based on Hyland’s MD model, the analysis was conducted using ANTCONC, a corpus analytical software. Compared to existing research on MD, the findings indicate that Arab L2 writers tend to use limited MD resources in their writing. The findings also reveal that there is a lack of balance in the use of interactional MD types in the corpus. Some interactional MD types are heavily used while some others are scarce. The discussion section of RAs received the highest frequency of interactional MD whereas the methods sections received the lowest. Based on these findings, some pedagogical implications for developing interactional competence of Arab L2 writers are discussed.