How Does That Large Language Model Make You Feel?
Clinical Snapshot
PICO Framework
| P — Population | General public and individuals seeking emotional support from commercially available large language models (LLMs) |
| I — Intervention | Use of commercially available large language models (LLMs) for emotional support and mental health purposes |
| C — Comparator | Not formally defined; implicit comparison to conventional mental health support or no support |
| O — Outcomes | Safety concerns, research gaps, and expert perspectives on the role of LLMs in mental health support |
Bottom Line
This News and Perspectives piece from JMIR raises important questions about the growing use of commercially available large language models for emotional support, drawing on expert commentary to highlight safety concerns and research gaps. However, it provides no original empirical data and represents the lowest tier of clinical evidence. Clinicians should treat its conclusions as hypothesis-generating rather than practice-changing. The core clinical message — that patients are increasingly using LLMs for mental health support in an evidence-free, largely unregulated environment — is plausible and warrants attention. In Australian practice, where mental health workforce shortages drive patients toward digital alternatives, this phenomenon is particularly salient. Practitioners are advised to proactively ask patients about LLM use as part of mental health histories, counsel on limitations and potential harms (including crisis escalation risks and misinformation), and direct patients toward evidence-based digital mental health resources such as MindSpot, This Way Up, or regulated telehealth services. Robust prospective research, including randomised controlled trials and longitudinal safety studies, is urgently needed before LLMs can be endorsed for any therapeutic mental health role.
Key Findings
P Value: Not reported
Effect Size: Not applicable — no empirical data presented
Primary Outcome: Expert perspectives on the safety, research gaps, and future directions for LLM use in emotional support and mental health — no quantitative primary outcome reported
Nnt Or Sensitivity: Not applicable — commentary format; no NNT, sensitivity, specificity, or hazard ratio calculable
Confidence Interval: Not reported
Clinical Application
The article does not propose a specific clinical intervention. Its value lies in awareness-raising for clinicians about patient behaviours and emerging safety concerns. No implementation pathway, resource requirement, or workflow change is specified. Clinicians can feasibly incorporate awareness of LLM use into routine mental health screening conversations. Highly relevant to Australian practice given the rapid uptake of LLM tools (e.g., ChatGPT, Gemini, Claude) among the Australian public. The TGA does not currently regulate general-purpose LLMs as medical devices unless they make specific diagnostic or therapeutic claims, creating a regulatory gap. The RACGP has not yet issued formal guidance on LLM use in mental health support as of mid-2025. Medicare does not fund LLM-based mental health interventions, and PBS listing is not applicable. Australian clinicians should be aware that patients in rural and remote areas, where access to mental health professionals is limited, may be particularly likely to turn to LLMs as a first-line resource. Headspace, Beyond Blue, and Lifeline digital services represent regulated alternatives. The Australian Digital Health Agency's ongoing work on AI governance is pertinent context. Clinicians across primary care, psychiatry, psychology, and digital health who encounter patients using or considering LLMs for emotional support; relevant to any practitioner whose patients have access to commercially available AI chatbots
Abstract
People are increasingly turning to commercially available large language models (LLMs) for emotional support. In this News and Perspectives article, JMIR Correspondent Simon Spichak reports on the role of LLMs in mental health, speaking with experts about safety concerns, research gaps, and next steps.
References
- 1.Spichak, S. (2026). How does that large language model make you feel? Journal of Medical Internet Research. https://doi.org/10.2196/105105
Related Research
Cognitive, affective & behavioral neuroscience
Emotion detection unveiled: A cognitive-computational synthesis of physiological models, machine learning, and datasets.
3 Aug 2026
Cognitive, affective & behavioral neuroscience
A review of current capabilities and future directions in machine-based emotion recognition
2 Aug 2026
JMIR mHealth and uHealth
Machine Learning Frameworks for Wearable-Based Stress Modeling in Naturalistic Settings: Scoping Review
1 Aug 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service