How students make sense of ChatGPT in digital literacy education: reflections on ethics, dependency, emotion, and evaluation

Generative AI tools such as ChatGPT are rapidly entering higher education and reshaping how students access, justify, and evaluate knowledge, issues relevant to academic integrity and information governance. However, most studies focus on adoption, attitudes, or performance outcomes, and fewer examine how students reflectively make sense of ChatGPT after structured digital literacy instruction, including how ethics, dependence, emotion, and credibility evaluation intersect. This study analyzed 133 undergraduate self-reflection texts produced after a Digital Literacy course at an Indonesian university. We conducted a qualitative content analysis using NVivo, combining prompt-guided coding with iterative thematic abstraction to identify higher-order patterns across reflections. Three higher-order themes emerged, negotiated dependency captured situational reliance on ChatGPT alongside efforts to retain authorship and learning autonomy. Ethical ambivalence described boundary-setting around fairness, originality, and disclosure. Conditional trust reflected trust calibration through verification practices (e.g., cross-checking, selective rewriting) and recognition that fluent outputs may be unreliable. Emotional responses, including reassurance, anxiety, and guilt, often shaped the intensity of reliance and the likelihood of verification. Students did not uniformly treat ChatGPT as an epistemic authority; instead, they calibrated their responsibility and trust in the face of uncertainty. The study’s novelty lies in conceptualizing AI literacy as “layered positioning” across dependency, ethics, affect, and evaluative judgment, showing how these dimensions co-occur in students’ narratives rather than operate separately. The findings support policy and curriculum design that make disclosure expectations explicit, teach verification routines, and use reflective assessment to strengthen student agency in AI-mediated knowledge environments.