Generative artificial intelligence use patterns among Chinese nursing graduate students: A qualitative descriptive study.
TL;DR
To explore how nursing graduate students in China experience and navigate generative artificial intelligence use in their research practice, including the conditions under which such use becomes dependency, a differentiated, stage-sensitive artificial intelligence governance frameworks tailored to nursing graduate education is advocated.
Abstract
Background
Generative artificial intelligence is increasingly embedded in graduate nursing research environments. Within this context, patterns of reliance on these tools, extending in some cases to dependency, have emerged among nursing graduate students, raising concerns about the impact on the development of foundational research competencies. Yet, limited qualitative evidence exists regarding how such reliance manifests, evolves, and is experienced by students at different training stages.
Objectives
To explore how nursing graduate students in China experience and navigate generative artificial intelligence use in their research practice, including the conditions under which such use becomes dependency.
Design
Qualitative descriptive study.
Setting
Data were collected from nursing graduate programs at 10 universities across China.
Participants
Seventeen nursing graduate students (9 doctoral, 8 master's) were recruited using purposive sampling supplemented by snowball sampling. All participants had documented use of generative artificial intelligence tools for research-related tasks for a minimum of three months.
Methods
Semi-structured in-depth interviews were conducted between March and April 2026. Data were analyzed using Braun and Clarke's six-phase reflexive thematic analysis, supported by NVivo 12.0 software, and interpreted through cognitive offloading theory.
Results
Data analysis identified four themes: (1) pervasive generative artificial intelligence integration across research workflows; (2) how generative artificial intelligence simultaneously augments output and erodes capability; (3) emotional and evaluative responses to generative artificial intelligence dependency; and (4) self-regulation strategies and structural gaps. Cross-case analysis yielded a typology of four user profiles: tool-rational, collaborative self-regulating, efficiency-oriented explorer, and deeply dependent user. Doctoral students predominantly occupied profiles characterized by higher metacognitive control, suggesting an "experience-buffering effect" from prior research training.
Conclusions
Generative artificial intelligence integration among nursing graduate students simultaneously enhances research productivity while risking the covert erosion of foundational competencies. The central challenge shifts from prohibiting artificial intelligence use to ensuring that metacognitive control keeps pace with integration depth. These findings advocate for the implementation of differentiated, stage-sensitive artificial intelligence governance frameworks tailored to nursing graduate education.