Applying Andersen’s Behavioral Model to Traditional Therapy Use and Generative AI Use for Mental Health Among College Students

Background and Purpose: Mental health concerns are widespread among college students, yet many students with need do not access formal counseling or therapy because of barriers such as stigma, cost, wait times, and limited service capacity (Cadigan et al., 2019; Fullmer et al., 2021). Generative AI tools are also emerging as a source of mental health information and support. Although Andersen’s Behavioral Model of Health Services Use has been widely used to explain formal service utilization, less is known about whether it also helps explain generative AI use for mental health purposes (Andersen, 1995). This framework may clarify whether AI-based mental health use reflects need-driven help-seeking like therapy or a distinct modality shaped by enabling beliefs such as trust in AI. This study applied Andersen’s model to compare predictors of current therapy use and generative AI use for mental health among college students.

Methods: Data came from the 2024–2025 Healthy Minds Study, a web-based survey of students at participating U.S. colleges and universities. Analyses were limited to respondents who completed the elective AI module (N=953). Two binary outcomes were examined: current therapy use and any generative AI use for mental health. Predictors were organized according to Andersen’s Behavioral Model as predisposing factors (sociodemographic characteristics), enabling factors (trust in AI, stigma, and financial stress), and need factors (mental health symptoms, perceived need, and academic impairment). Hierarchical logistic regression models were estimated separately, with Model 1 including predisposing factors, Model 2 adding enabling factors, and Model 3 adding need factors.

Results: In the analytic sample, 26.2% of respondents reported current counseling or therapy use, and 14.8% reported any generative AI use for mental health. In the fully adjusted therapy model, perceived need for mental health help was the strongest correlate of therapy use (OR=2.690, p<.001). Age (OR=1.069, p=.041) and gender-diverse identity (OR=2.447, p=.001) were also associated with higher odds of therapy use. In contrast, AI use for mental health was most strongly associated with enabling beliefs. Trust in AI for mental health information (OR=2.730, p<.001) and trust in AI for mental health decisions (OR=1.827, p=.002) were the strongest predictors. Asian students had higher odds of AI-for-mental-health use (OR=1.897, p=.015), and perceived need was positively associated (OR=1.186, p=.035). Depressive symptoms, anxiety symptoms, and loneliness were not significant predictors in the fully adjusted AI model.

Conclusions and Implications: Findings suggest that generative AI use for mental health may represent a distinct help-seeking modality rather than simply a digital extension of traditional therapy. Whereas therapy use was primarily need-driven, AI-based mental health use was more strongly shaped by enabling beliefs, especially trust in AI. For social work practice and campus mental health systems, these findings underscore the need to recognize AI as an emerging part of the student help-seeking landscape. Institutions may need to address unmet mental health need and students’ beliefs about the credibility, usefulness, and limitations of AI-based support. Andersen’s model may remain useful for understanding these behaviors, but emerging AI-based help-seeking may require greater attention to technology-specific enabling factors.