Young adults of color say ChatGPT mental health advice overlooks their cultural experiences

Research involving 13 young adults of color identifies concerns around generic AI advice, personal data disclosure and unclear boundaries between chatbot support and clinical care

New research examines how young adults of color experience AI-generated mental health support, including concerns about cultural context, personalization and privacy

Young adults seeking mental health support from ChatGPT found that its responses sometimes overlooked the cultural and personal circumstances shaping their concerns, according to new US research examining how AI chatbots respond to users from different backgrounds.

The study, led by researchers at the University of Southern California, involved 13 Asian, Black and Hispanic or Latino young adults aged 18 to 24. Participants took part in a two-day workshop, testing ChatGPT with mental health-related scenarios before designing features they believed would make AI support more appropriate and responsive to their needs.

Their experiences revealed a tension between personalization and privacy. While participants wanted AI to understand more about their identities, family relationships and cultural backgrounds, relatively few initially considered the potential risks of disclosing extensive personal information.

The findings appear in a research preprint titled Co-Designing AI For Mental Health Support With Young Adults of Color: Needs, Expectations, and Implications for AI Literacy, published on October 2, 2026. The paper has not yet been peer reviewed.

Generic responses overlooked family pressure and cultural experiences

During the workshop, participants explored scenarios involving academic stress, social anxiety, parental expectations, immigration experiences and interpersonal relationships.

Although ChatGPT could provide practical suggestions, participants reported that its responses sometimes treated culturally influenced experiences as individual problems.

One Asian international student, seeking guidance on academic stress and language barriers at an American university, observed that ChatGPT "didn’t ask about my background, so the advice was somewhat general".

Another participant described how the chatbot responded to burnout associated with parental sacrifice and family obligations by recommending self-care, without adequately recognizing the cultural expectations involved.

The issue extended beyond insufficient personalization. Participants also identified occasions when ChatGPT made incorrect assumptions about their circumstances, including treating an international student's experience as equivalent to that of an immigrant establishing a permanent life in another country.

Rather than expecting chatbots to infer their identities, participants wanted more intentional follow-up questions and opportunities to explain relevant aspects of their backgrounds themselves.

Greater personalization raised questions about privacy

The desire for more contextualized responses produced another finding: participants were frequently willing to disclose substantial personal information if they believed it would improve the advice they received.

Some proposed introductory questionnaires collecting details such as ethnicity, immigration history, household circumstances and formative experiences. They wanted this information to help shape subsequent conversations without repeatedly having to explain themselves.

However, discussions about privacy were comparatively limited. Only one participant explicitly identified personal privacy breaches as a concern in the initial survey.

Researchers found that participants often associated privacy risks with other people sharing sensitive information, rather than considering how AI systems might collect, retain or use personal data over repeated interactions.

The workshop itself used university-contracted ChatGPT EDU accounts and temporary chats to protect participants' information.

A separate concern emerged around the boundaries of AI mental health support. Participants described instances where ChatGPT initially offered therapy-style guidance before later stating its limitations and directing users toward crisis resources or professional support.

They wanted those boundaries established earlier, rather than discovering them after disclosing sensitive information.

In response, participants developed several proposed features, including clearer explanations of chatbot capabilities, user-controlled conversation pathways and activity summaries designed to encourage reflection on AI use. These were workshop design concepts, not implemented product changes.

Sharing the findings on LinkedIn, Michael Robb, Head of Research at Common Sense Media and its Youth AI Safety Institute, wrote:

"AI literacy for young people needs to cover privacy and what a chatbot can and can’t do for your mental health."

The researchers argue that AI literacy education should be developed alongside mental health literacy, helping young adults understand not only how chatbots operate, but also their limitations, privacy implications and differences from professional mental health care.

The study was conducted during summer 2026 and involved a small group from one metropolitan area on the US West Coast. Its findings are not intended to represent the experiences of all young adults of color, and the authors acknowledge that the facilitated workshop setting may have influenced participants' responses.

The research team plans to adapt its pilot workshop approach for other groups, including younger teens and people living with mental health diagnoses.

ETIH Innovation Awards 2027 open for entries

The ETIH Innovation Awards 2027 are now open for entries, with 26 categories recognising companies, institutions, projects and individuals making a measurable impact across education technology. The global awards span AI in education, digital learning, higher education, skills and workforce, inclusion, education operations, evidence and impact, with entries managed through ETIH’s new online awards platform.

Entries are open worldwide, with the early-bird deadline on Friday 18 December 2026 and the final entry deadline on Friday 26 February 2027 at 23:59 UK time. 

Explore the 2027 categories and judging criteria, read the How to Enter guide or start an entry through the ETIH Innovation Awards 2027 platform.