“I Feel More Comfortable Speaking to an AI, Not an Actual Person” – Aanhpi Young Adults and Their Use of AI for Mental Health

Background and Purpose:

Asian American, Native Hawaiian, and Pacific Islander (AANHPI) young adults (ages 18–29) experience psychological distress at rates comparable to or greater than other U.S. racial/ethnic groups, yet significantly underutilize mental health services. Cultural norms (e.g., stoicism, filial piety), systemic barriers, and the "model minority" stereotype often discourage formal help-seeking. Artificial Intelligence (AI) chatbots (e.g., ChatGPT) and Large Language Model (LLM)-based tools are touted as accessible avenues that can bypass these barriers. However, little is known about how AANHPI young adults integrate AI into their mental health journeys. To address this gap, this study qualitatively explores individual and social factors shaping the use of AI to support mental health among AANHPI young adults.

Methods:

We conducted semi-structured 60-minute interviews with a diverse sample of 38 AANHPI young adults. The interview protocol drew from the Seeking Mental Health Care model and the Network Episode Model, and explored participants' illness experiences, social network influences, and digital treatment navigation. Transcripts were analyzed in Dedoose using a reflexive thematic analysis. We developed a codebook to capture the dynamic interplay between symptom recognition, social influence, and specific digital modalities. This analysis isolated codes related to AI tools to understand user experience, AI-prompting, sociocultural and network factors, and the perceived benefits and limitations of non-human mental health support.

Results:

Several key themes emerged. First, users preferred AI over traditional therapy and human interaction due to the anonymity it provides. One participant noted that communicating with AI is "very secret," allowing them to bypass the cultural shame or "loss of face" associated with visible, formal care. Second, heavier AI users often reported previous negative, invalidating, or unsuccessful experiences with human therapists, prompting their turning to AI.

Conceptions of what constitutes mental health support varied. Some participants adopted a broad definition, utilizing AI for practical, stress-reducing solutions (e.g., drafting difficult interpersonal messages or navigating conflict). Others viewed support through a narrower lens of direct emotional care. Furthermore, users noted that higher mental health literacy enabled more effective prompting, resulting in highly tailored support.

Finally, general sentiment toward AI heavily influenced engagement. Participants who perceived AI negatively were significantly less likely to use it, citing environmental impacts, plagiarism, social stigma, and privacy concerns. Notably, most participants relied on generic AI chatbots (e.g., ChatGPT, Gemini, Claude) while largely avoiding specialized mental health AI tools.

Conclusions and Implications:

AI chatbots are one promising but imperfect avenue of mental health support for AANHPI young adults. Rather than simply acting as an "on-ramp" for the unengaged, AI seems to best serve as a private refuge for individuals failed by traditional therapeutic systems and those possessing the mental health literacy to guide their own care. However, evolving sentiment around AI can turn away certain users. Social work practitioners and policymakers should recognize AI platforms as valid support options within the AANHPI community, meeting clients where they are and providing appropriate education rather than dismissing these tools.