Metadiscourse features are rhetorical devices deployed by authors to establish interaction with prospective readers. These devices do not contribute to the propositional content of discourse. Considering this premise, the present study aimed to examine the distributional patterns of metadiscourse features in academic discourse in English and Persian, based on Hyland’s classification. To achieve this goal, a do-it-yourself (DIY) corpus comprising 1,570,786 tokens was compiled from 273 texts in English and Persian. The data were analyzed using the Sketch Engine corpus software. The results revealed that the Persian corpus contained a higher number of metadiscourse features compared to the English corpus. Furthermore, while the Persian corpus exhibited a preference for interactive features, the English corpus showed a stronger inclination toward interactional ones. Additionally, a close reading of the concordance lines indicated significant differences in how writer–reader interaction is constructed in English and Persian. The findings of this study are expected to have practical implications for fields such as contrastive linguistics, corpus linguistics, genre analysis, and contrastive analysis.