Brazilian and French University Students’ perceptions of the quality of text generated by a GenAI Chatbot

Abstract

From a literacy studies perspective, this article analyzes, in the responses of Brazilian and French university students, perceptions of the quality of texts produced by a generative artificial intelligence (GenAI) tool in an academic context. The pedagogical activity that guided text production explicitly considered the potential use of AI-based chatbots in university coursework. In analysing these perceptions, the study aims (i) to identify the linguistic resources mobilised by Brazilian and French students when assessing the credibility of GenAI-generated texts for academic purposes, and (ii) to discuss how these resources inform understandings of text quality across different contexts. The corpus comprises 92 written responses produced by undergraduate students enrolled in teacher-education programmes in both countries (Brazil and France). Drawing on a mixed-methods approach, the analysis articulates (i) degrees of trust with (ii) two sets of linguistic criteria: academic-normative features and semantic-argumentative features. The findings reveal a predominance of qualified or partial trust, with evaluations largely grounded in academic-normative criteria such as formal correctness and genre conventions. Semantic-argumentative features, related to the quality and coherence of argumentation, emerge only sporadically, particularly in the Brazilian dataset. These results suggest that the primacy of academic norms may frame students’ judgements of text quality, potentially restricting critical reading practices to a focus on surface-level linguistic features. The article argues for pedagogical approaches to digital technologies that integrate both dimensions, thereby strengthening discursive and critical literacies essential for addressing misinformation in higher education.