Whose English counts? Linguistic justice, gatekeeping, and AI-mediated language policy in academic publishing

Academic publishing has long served as a site of language regulation, shaping which knowledge becomes publishable, under what linguistic conditions, and at what cost. The normalization of generative artificial intelligence sharpens these concerns, as AI tools are now increasingly entangled with manuscript preparation, editorial screening, and peer-review governance. This article examines AI-mediated academic publishing as an emerging domain of language policy and planning. It reports a qualitative document analysis of 10 author-facing policy texts from six major publishing ecosystems: Elsevier, Springer Nature, Wiley, Taylor & Francis, SAGE, and Oxford University Press/Oxford Academic. Combining structured qualitative content coding with reflexive thematic analysis, the study examines how these documents regulate language assistance, authorship, disclosure, responsibility, and textual acceptability. The analysis identifies four recurring policy logics: conditional permission for AI-assisted language improvement, prohibition of AI authorship, individualized author responsibility, and uneven disclosure thresholds. The article argues that AI does not simply lower linguistic barriers to publication. Rather, it reworks standard language ideology, redistributes compliance burdens, and subjects multilingual scholars to new forms of suspicion, disclosure risk, and credibility assessment within the infrastructures of scholarly publishing.