A four-profile latent structure identified among Chinese university students showed that GenAI-related digital behavior co-occurred with dysregulated use, academic anxiety, sleep problems, distress, and academic functioning impairment, while higher health literacy and self-regulated learning were associated with greater odds of low-risk rather than high-risk membership.
Latent profile analysis identified four distinct GenAI-use and mental-health risk profiles among Chinese university students.
Higher health literacy and self-regulated learning were associated with greater odds of belonging to the low-risk rather than high-risk profile.
Network analysis identified loss of control, academic worry, delayed bedtime, and academic avoidance as central or bridge nodes in the symptom network.
The measurement structure of adapted and selected scales was supported by confirmatory factor analysis with good fit indices.
The study retained 1,872 valid responses from 2,140 submitted questionnaires after a four-step data-quality screening process.
GenAI-related problematic use co-occurred with multiple mental health and academic functioning concerns including anxiety, sleep problems, psychological distress, and impaired academic functioning.
Credibility evaluation and time management showed negative bridge expected influence in the network, suggesting these may function as protective factors.
This research suggests that Chinese university students are not a uniform group when it comes to how they use generative AI tools like ChatGPT and whether that use causes harm. By surveying nearly 1,900 students across six universities, researchers found four distinct types of users: those who use AI in a healthy, controlled way (about 40% of students), those who experience heightened anxiety tied to their AI use (about 26%), those whose AI use disrupts their sleep (about 20%), and a high-risk group who have difficulty controlling their use and experience multiple mental health problems simultaneously (about 15%). The fact that one in seven students fell into the high-risk group suggests that problematic AI use is a meaningful public health concern on university campuses. The study also identified which specific symptoms or behaviors seem to be most central in linking these problems together. Loss of control over AI use, worrying about academics, going to bed later than intended, and avoiding academic tasks were the key 'hub' problems — meaning addressing these might have the most impact across multiple areas of student wellbeing. On the protective side, the ability to evaluate whether AI-generated information is credible and the ability to manage one's time appeared to act as buffers against risk spreading through the network of problems. This research suggests that universities could focus on building students' health literacy — meaning their ability to find, understand, and critically evaluate health and information sources — alongside self-regulation and time management skills as practical ways to help students avoid problematic AI use patterns. The researchers note these are only associations from a single snapshot in time, so future studies following students over time and testing actual interventions are needed to confirm whether improving these skills genuinely reduces harm.
Zhang Z, Zhang Y, Lu X, Zhang M. (2026). Problematic generative AI use and mental health risks among Chinese university students: latent profiles, network structure, and health literacy as a modifiable resource.. Frontiers in public health. https://doi.org/10.3389/fpubh.2026.1910408