The Relationship between Latent Profiles of Psychosocial Vulnerability and AI Counseling Utilization Among College Students

Background and Purpose: College students use AI tools for a variety of purposes, including academic tasks, information seeking, and writing support. More recently, there has been a growing demand for AI-based interactions in psychological domains such as emotional concerns and stress. As more college students than ever use generative AI for counseling, concerns have arisen in regard to the potential risks. Large language models may distort users’ judgment, potentially leading to negative and unethical outcomes, particularly among college students with mental health vulnerabilities. Therefore, to address these potential risks, it is critical to identify the characteristics of users who choose AI counseling. Given the benefits and risks of AI-based counseling, researchers have paid little attention to how college students’ psychosocial characteristics are associated with use of digital platforms. Thus, this study explores college students’ use of AI-based counseling, focusing on psychological vulnerabilities and the availability of coping resources.

Methods: This cross-sectional study recruited college students aged 19 to 24 across South Korea to ensure the representativeness of the samples. The study was conducted through an online survey from November 2025 to December 2025. For analysis, 514 participants completed the survey. This study used latent profile analysis to identify distinct patterns based on psychological vulnerabilities such as depression, anxiety, and stress, and available coping resources, such as problem-solving ability and social support. Binary logistic regression analysis was conducted to examine the effects of the identified latent profiles on the participants’ utilization of AI counseling. Setting the latent profile groups as independent variables, with the lowest vulnerability group set at the reference group, allowed comparison of moderate and high vulnerability groups’ relative likelihood of using AI counseling.

Results: Among the 514 participants, 259 (50.4%) were male and 255 (49.6%) were female. The mean age was 21.51 (SD=1.55). A total of 270 participants (52.5%) reported having used AI for psychological counseling within the past six months; 244 (47.5%) reported that they had not. Binary logistic regression analysis showed that the model was statistically significant (χ²=18.612, p=.017), and the Hosmer–Lemeshow test was non-significant (p=.347), which indicates a good model fit. The results show that compared to the low vulnerability group with low psychological vulnerabilities and high coping resources, the moderate vulnerability group and high vulnerability group were 1.51 times (B=.41, p<.05) and 2.62 times (B=.96, p<.001), more likely to use AI-based counseling, respectively.

Conclusion and Implications: These findings suggest that college students with the highest levels of psychological vulnerabilities and lower coping resources, had a greater need for psychological support and were therefore more likely to take advantage of AI-based counseling. This result indicates that AI counseling may serve as a supplementary pathway for early help-seeking, particularly among college students with limited access to traditional mental health services. These findings suggest that clinical social workers should assess college students’ psychological vulnerabilities and guide the safe, ethical, and tailored use of AI-based counseling while also providing traditional mental health services for students at elevated risk of mental vulnerability.