Helpful for whom? Rethinking youth preference for AI in mental health advice

To the Editor,

Skjuve et al. provide a rigorous comparative evaluation showing that young people rated ChatGPT’s responses to mental health inquiries as more relevant and useful than responses written by health professionals. ¹ Their study is timely and relevant given the increasing reliance of young people on digital spaces for low-threshold mental health information and support. ² Yet the category of “youth” in this discussion requires greater conceptual caution. Rather than implying a uniform youth experience, these findings are better understood as patterns observed across the group, which may still vary according to young people’s contexts, vulnerabilities, and available support systems.

This is particularly important when seen from the Philippine perspective, where child and adolescent mental health support remains marked by unmet need, delayed access to care, fragmented implementation, and limited community-level responsiveness. ³ In such a setting, it is unsurprising that immediate, structured, and actionable answers are perceived as helpful. However, one must be careful not to interpret this perceived helpfulness as a form of superior mental health support. It may also reflect what is communicatively accessible in a setting where professional systems remain uneven and difficult to reach. Cornelio’s caution against overgeneralized generational narratives then becomes a useful interpretive lens. ⁴ Rather than treating Filipino youth as a homogeneous category, his perspective foregrounds fragmentation, inequality, and historically situated vulnerability. Applied to Skjuve et al.’s findings, this suggests that “youth preference” for ChatGPT may not simply indicate better advice. It may also reflect how particular groups of young people respond to communicative features that feel fluent, scaffolded, immediate, and usable under conditions of precarity. In this sense, preference may be less about therapeutic superiority than about social and communicative fit.

Indeed, Skjuve et al. show that answers were well received when concrete, well-structured, and easy to process. Conversely, professionals were more likely to find ChatGPT too detailed, too factual, or insufficiently aligned with established communication norms. ¹ This raises an important distinction. ChatGPT may function, in some cases, more as a better explainer than as a better adviser. For young people negotiating distress in unequal contexts, pedagogical clarity may easily be experienced as care itself. Still, ease of access and perceived usefulness should not be equated with safety. AI-generated responses, however clear or immediate, cannot replace professional judgment, safeguarding, and appropriate referral pathways, particularly in complex or high-risk mental health situations.

Future research should therefore move beyond generational shorthand and ask which young people find AI advice helpful, under which conditions of vulnerability, and in response to what gaps in existing support. Otherwise, youth preference for AI may obscure the unequal social worlds through which young people encounter mental health care. The implication is not merely methodological but systemic. If young people value AI-generated advice for its immediacy, clarity, and low-threshold character, this appeal may reveal both the promise of digital tools and the continuing need to make formal mental health systems more accessible, usable, and responsive to safety concerns. In this sense, the findings of Skjuve et al. do not simply show what ChatGPT can provide, but more importantly, invite further reflection on how mental health systems can make care feel more legible, readable, and reachable to the young people who need it.