Digital translation practices in university English learning: adoption, disciplinary variation, and AI-assisted revision

Introduction Machine translation and generative artificial intelligence (AI) are now common resources in English as a foreign language learning, yet evidence from Latin American higher education remains limited and often conflates tool use with learning impact. This study characterized translation practices, tool preferences, print-dictionary use, and AI-assisted text revision among university students in southern Peru. Methods A cross-sectional survey was conducted with 435 students from 37 degree programs and seven faculties at a public university. Questionnaire items were analyzed as separate indicators rather than combined into a latent scale. Valid percentages with 95% Wilson confidence intervals were estimated, paired exact McNemar tests were used for within-participant comparisons, faculty differences were assessed using Monte Carlo chi-square tests with false-discovery-rate adjustment, and an exploratory logistic regression was fitted. Results Use of AI or automated translators was reported by 94.7% of respondents, and 92.4% used translation applications or websites. Both were more prevalent than print-dictionary experience (65.4%; paired odds ratios 9.86 and 7.88, respectively; both p < .001). Google Translate predominated across questionnaire contexts, whereas ChatGPT accounted for 7.7% of responses to the general-use item and 34.0% to the AI-focused item. AI-assisted revision was reported by 74.1% and varied across faculties (52.2%-96.8%; Cramér's V = .274, q < .001). Although 75.0% reported using print dictionaries rarely or never, 78.0% supported their continued promotion. Discussion The findings describe a heterogeneous institutional translation ecology in which widespread digital-tool adoption coexists with disciplinary variation, context-dependent tool preferences, and continued endorsement of print resources. These findings do not establish learning effects. They support pedagogical attention to verification, post-editing, and independent language production, whose effectiveness requires direct evaluation.