Beyond Polishing: The Compounding Dynamic of GenAI in Academic Writing

Abstract Generative AI can reshape stance and voice in multilingual student writing, with effects that appear to be shifting over time. Using a random sample of English L2 master’s-level assignments at a UK university from two cohorts (2019 and 2025, N = 25 each), I first compared submitted drafts, then subjected parallel excerpts to GenAI “polishing” prompts. Across both cohorts, the technology consistently reduced hedging, suppressed self-mention, and reframed authorial stance into more categorical and impersonal forms. While these outcomes confirm existing critiques of homogenisation, the more revealing finding is a compounding dynamic. I use this term within an ecological view of writing and technological mediation to describe how AI-mediated drafting and AI-based polishing interact in the same writing environment. The 2025 students’ original texts, acknowledged by most to have been written with GenAI support, already resembled AI outputs, and polishing then reinforced the drift. Compared to the 2019 cohort, their drafts contained fewer hedges, greater lexical diversity, and syntactically simpler sentences. This longitudinal perspective shows how the stylistic fingerprints of GenAI are being internalised within writing practices, narrowing rhetorical space even before the tool is explicitly applied. I highlight the pedagogical imperative of critical AI literacy: that writers must be trained to use generative AI effectively by interrogating its outputs, resisting its homogenising pull, and preserving their intended stance.