Translating Power: AI, Mediation and the Imaginary of Manipulation in Academic Contexts

The growing use of generative artificial intelligence in academic writing and translation is changing not only the speed of textual production but also the ways in which linguistic authority is constructed and perceived. In multilingual academic contexts, AI systems translate, reformulate, and refine texts according to dominant conventions of academic English. Such assistance may widen access to international communication, yet its fluency can also make the interpretative choices embedded in the resulting text less visible. This article examines AI-mediated Romanian-English academic translation as a site where linguistic assistance, symbolic power, and the imaginary of manipulation intersect. The study approaches translation as mediation rather than neutral transfer. It asks how AI may reshape lexical authority, epistemic stance, and stylistic identity, how these changes influence the perceived legitimacy of academic discourse, and when unobtrusive mediation may acquire a potentially manipulative effect. The theoretical discussion connects translation studies with scholarship on symbolic power, academic stance, and sociotechnical imaginaries. Methodologically, the article adopts a conceptual qualitative approach supported by three illustrative Romanian-English micro-analyses. These examples are not treated as a representative corpus and do not support statistical generalisation. Instead, they reveal how apparently minor lexical, epistemic, and stylistic reformulations can influence authority, certainty, and authorial visibility. The article argues that manipulation should not be equated automatically with factual distortion or deliberate ideological intent. It may also arise through naturalisation, when a fluent and plausible formulation is accepted as transparent or inevitable. The article concludes that responsible AI-assisted academic translation requires comparison with the source, attention to shifts in certainty, agency and rhetorical positioning, and clear human responsibility for the final text. For translators, writers and educators, critical AI literacy entails treating generated language as a proposal to be evaluated rather than as an impersonal solution.