Based on an interview with Yasaman Kambari
As artificial intelligence becomes increasingly integrated into health care, Yasaman Kambari has been watching that shift unfold alongside her own training. Kambari is a graduate student in Clinical Neuropsychology at York University who is training for a career in mental health care. During her training, she began hearing clinicians talk more frequently about using AI for tasks such as summarizing clinical notes. She was also struck by how quickly AI was becoming part of health-care encounters more broadly, with some people now being given the option to consent to AI-assisted tools being used during their visits.
For Kambari, these changes raised an important set of questions. Who is comfortable with AI becoming part of mental health care? Who is not? And why? As someone preparing to work in this field, she was interested not only in what AI could do, but in how people might respond to its adoption. “It’s helpful to ask, ‘Who will this benefit?’” she questioned.
That question became particularly important when she considered the role of gender. Mental health concerns, experiences with care, and patterns of help-seeking can differ across gender identities. If those differences already exist in traditional mental health systems, what happens when AI enters the picture?
When Access is Not Enough
AI’s appeal in mental health care is clear: it promises support that is more immediate, accessible, and affordable. A digital tool can be available at any hour, including when a clinician is not. It may offer lower-cost support, help someone access resources before reaching a crisis point, or provide a greater sense of anonymity for people who are uncomfortable seeking more traditional forms of care. For mental health systems struggling with limited resources and long waitlists, these possibilities are meaningful.
But making support available does not necessarily mean making it equitable. Digital mental health tools still require a device, internet access, and some degree of digital literacy. And even when a tool is technically available, people may differ greatly in whether they trust it or want to use it. As Kambari put it, “Increasing access doesn’t really mean that everyone will use it, and everyone will benefit from it.” That distinction between access and actual benefit is central to her work. A mental health tool may be inexpensive, immediate, and widely available, yet still leave some of the people it is intended to support behind.
Why Gender Matters
Gender can shape how people experience mental health, seek support, and interact with health-care systems. These differences are especially important when considering gender-diverse people, who may face additional barriers to finding care that feels affirming and responsive to their experiences. Cisgender describes people whose gender aligns with the sex they were assigned at birth. Gender-diverse is a broad term for people whose gender identities or expressions differ from conventional expectations associated with the sex they were assigned at birth; non-binary identities are one form of gender diversity.
Research has found that non-binary youth report poorer general mental health and higher symptoms of depression and anxiety than cisgender youth. Experiences of gender-identity invalidation, in which a person’s gender is dismissed or treated as illegitimate, have also been linked with poorer mental health among non-binary people (Johnson et al., 2024; Klinger et al., 2024). These disparities do not arise from gender diversity itself. They exist within a broader context of social marginalization, discrimination, and barriers to affirming care. That context becomes especially important when a new technology is introduced into mental health care.
What Changes When AI Enters the Picture?
Kambari, along with her advisor Dr. Achala H. Rodrigo, examined this question in a study of 1,025 Canadian adults, including women, men, and non-binary participants. Participants were asked about their willingness to use mental health services delivered by a human clinician, by AI working collaboratively with a clinician, or by AI alone.
As AI took on a greater hypothetical role in delivering mental health care, acceptance increasingly differed across gender identities. When mental health services were described as being delivered by a human clinician, there were no significant differences in acceptance across gender identities. When services were described as being delivered collaboratively by AI and a clinician, non-binary participants reported lower acceptance than men and women. When services were described as AI-delivered, the differences became broader, with women reporting the highest acceptance, followed by men and then non-binary participants.
Kambari’s study cannot tell us exactly why these differences appeared. But previous experiences with health care can shape whether a new form of care feels trustworthy. For transgender and gender-diverse people, those experiences may include discrimination, invalidation, or difficulty finding providers who understand and affirm their identities. Research has linked such experiences with greater medical mistrust and, in some cases, avoiding needed care (Mason et al., 2024). Research with sexual- and gender-minority young adults has also found that transgender and gender-diverse participants were more likely than cisgender sexual-minority participants to anticipate discrimination in health-care settings. Across participants, greater trust in health-care providers was associated with lower expectations of discrimination (Snyder et al., 2026).
These studies do not explain Kambari’s findings directly, but they show why trust, representation, previous experiences of invalidation, and concerns such as privacy are important questions for future research. For Kambari, this makes it important to ask not only why some communities may be less willing to use AI-based mental health care, but whether the technologies themselves have been designed in ways that deserve their trust, questions that her current research aims to address, supported by a graduate student research award from the Connected Minds CFREF initiative.
Who Gets Represented?
Whether an AI system deserves people’s trust depends in part on how it was built. AI systems learn from existing data, so what they learn depends partly on whose experiences are represented and whose are missing.
Kambari emphasized that this matters especially in mental health care. If the data used to train an AI system primarily reflect certain populations, its outputs may not work equally well for everyone. For a client seeking mental health support, that could mean an inaccurate risk assessment, an inappropriate recommendation, or resources that do not fit their experiences or needs. If someone repeatedly feels misunderstood or poorly represented by a tool, a service intended to increase access could instead deepen mistrust.
This raises several questions: Who was involved in designing and developing the system? Who was represented in the data it learned from? Who participated in testing and evaluating whether it actually worked for the communities it was intended to serve?
Kambari argued that improving representation in training data is only part of the solution. Gender-diverse people should also be meaningfully involved in the development and testing of AI systems, including being asked whether those systems feel useful, supportive, and validating. In the interview, she emphasized the importance of testing these tools with gender-diverse users rather than assuming that a system developed primarily around cisgender populations will automatically meet their needs.
That approach reflects broader guidance on responsible health AI. The World Health Organization has emphasized that people who will use or be affected by AI systems should participate in their design and evaluation, and that equity and inclusiveness should be considered throughout development rather than added after a system has already been built (World Health Organization, 2021).
Equitable AI, then, requires more than building a tool and making it available to everyone. The communities expected to use these systems should have a meaningful voice throughout the process, from design and development to testing, implementation, and ongoing evaluation.
Where Does AI Belong in Mental Health Care?
A more practical question is not simply whether AI should be used in mental health care, but what roles it is best suited to take on.
Kambari’s research suggests that people may be more comfortable with AI for some tasks than others. Participants were generally more receptive to AI performing structured or relatively passive functions, such as monitoring symptoms or supporting self-guided activities, than taking on more active and individualized responsibilities such as diagnosis. Kambari noted that these more complex forms of care require a greater degree of individualized understanding and empathy. She also emphasized that some forms of mental health care depend on the relationship itself. Human clinicians can offer individualized understanding, empathy, and interpersonal connection in ways that remain difficult to reproduce through AI.
These questions are especially relevant as AI-based mental health and well-being tools become more widely available. Kambari emphasized several things worth considering before relying on such a tool: Is it grounded in evidence-based practice? Were mental health professionals involved in its development? Is it transparent about what it can and cannot provide? And could there be a mismatch between what someone needs and what the tool is designed to provide?
Research on these tools is growing, but the picture is still developing. A recent systematic review and meta-analysis found that generative-AI mental health chatbots produced small-to-moderate improvements in mental health outcomes on average, although results varied considerably across studies and populations (Zhang et al., 2025). Another recent review found that users often valued features such as convenience, personalization, and perceived empathy, while also concluding that many of these tools remain at an early stage of development (Olisaeloka et al., 2026).
Together, these findings suggest that the most useful role for AI may depend on what task AI is being asked to perform, how well the tool has been evaluated, and whether clients have meaningful choices about how they receive care.
What Might This Mean for Students?
The same questions about access, trust, and choice become especially relevant on university campuses. Students often face many of the barriers that AI-based services promise to address, including long waitlists, high costs, limited time, and periods of intense stress. For someone adjusting to university for the first time, having immediate access to support or information could be especially valuable.
Kambari, whose research interests also include exploring effective ways to improve the mental health of postsecondary students, sees potential for AI to be thoughtfully integrated into university mental health services. At the same time, she emphasized that campuses bring together students with widely varied backgrounds, experiences, and needs. Making the same tool available to everyone does not mean every student will trust it, feel represented by it, or want to use it. AI-based support should therefore add to the options available to students rather than become their only path to care.
The importance of maintaining human alternatives became especially clear when the conversation turned to peer support. Kambari noted that connecting with someone who shares aspects of a student’s background or lived experience can itself be comforting. AI might help organize or streamline peer-support programs, but it cannot replace that sense of shared experience.
For universities, thoughtful integration of AI may therefore be less about finding one technological solution and more about preserving multiple pathways to support: AI-based resources for students who find them useful, human services for those who prefer them, and combinations of the two where they complement one another.
Building AI with the People it Serves
The questions that first drew Kambari to this work remain unresolved: Who is comfortable with AI becoming part of mental health care? Who is not? And why? Her next steps focus on understanding what shapes trust and engagement with AI, particularly among gender-diverse people, and what those communities would need from these technologies to feel genuinely supported.
As AI becomes more integrated into health care, the goal cannot simply be to make it available as widely as possible. The larger challenge is to build systems with the people they are meant to serve, to evaluate whether those systems effectively provide support across different communities, and to preserve meaningful choice in how care is received.
For mental health care, that means asking not only what AI can do, but who it works for, who was involved in shaping it, and whether people can still choose the human support they may want and need.

Yasaman Kambari is a doctoral student in Clinical Psychology at York University, specializing in neuropsychology. Her research explores individual differences in attitudes toward AI in mental health care. She is especially interested in how individual differences shape trust and acceptance of AI-based mental health services.
References
Johnson, K. C., LeBlanc, A. J., Dolezal, C., Singh, A. A., & Bockting, W. O. (2024). Invalidation and mental health among nonbinary individuals. Psychology of Sexual Orientation and Gender Diversity, 11(3), 413–424. https://doi.org/10.1037/sgd0000621
Klinger, D., Oehlke, S.-M., Riedl, S., Eschbaum, K., Zesch, H. E., Karwautz, A., Plener, P. L., & Kothgassner, O. D. (2024). Mental health of non-binary youth: A systematic review and meta-analysis. Child and Adolescent Psychiatry and Mental Health, 18, Article 126. https://doi.org/10.1186/s13034-024-00822-z
Mason, K. L., Hood, K. B., Perrin, P. B., Belgrave, F. Z., Allison, K. W., & Coston, B. E. (2024). Direct and vicarious exposure to healthcare discrimination and erasure among transgender and gender independent individuals: Testing the indirect effect of mistrust in healthcare on utilization behaviors. Social Science & Medicine, 348, 116806. https://doi.org/10.1016/j.socscimed.2024.116806
Olisaeloka, L., Richardson, C. G., Wang, A. Y., Munthali, R. J., & Vigo, D. V. (2026). Generative AI mental health chatbots: A scoping review of intervention design and user experience. npj Digital Medicine. Advance online publication. https://doi.org/10.1038/s41746-026-02972-0
Snyder, J., Mason, K. L., & Hood, K. B. (2026). Trust in providers and social media stories on expectations of discrimination among sexual and gender minority young adults. Culture, Health & Sexuality, 28(7), 815–831. https://doi.org/10.1080/13691058.2025.2495752
World Health Organization. (2021). Ethics and governance of artificial intelligence for health: WHO guidance. https://www.who.int/publications/i/item/9789240029200
Zhang, Q., Zhang, R., Xiong, Y., Sui, Y., Tong, C., & Lin, F.-H. (2025). Generative AI mental health chatbots as therapeutic tools: Systematic review and meta-analysis of their role in reducing mental health issues. Journal of Medical Internet Research, 27, e78238. https://doi.org/10.2196/78238
