CoChat
Sign Up
Cited by 48Digital Mental Health InterventionsArtificial Intelligence in Healthcare and EducationEthics and Social Impacts of AI

Seeking Emotional and Mental Health Support From Generative AI: Mixed-Methods Study of ChatGPT User Experiences

Open access
Luo, Xiaochen, Wang, Zixuan, Tilley, Jacqueline L, Balarajan, Sanjeev, Bassey, Ukeme-Abasi, Cheang, Choi Ieng·2025·JMIR Mental Health
DOI 10.2196/77951PMC PMC12661908
Open this paper in CoChat — free account

Abstract

Background: Generative artificial intelligence (GenAI) models have emerged as a promising yet controversial tool for mental health. Objective: The purpose of this study is to understand the experiences of individuals who repeatedly used ChatGPT (GenAI) for emotional and mental health support (EMS). Methods: We recruited 270 adult participants across 29 countries who regularly used ChatGPT (OpenAI) for EMS during April 2024. Participants responded to quantitative survey questions on the frequency and helpfulness of using ChatGPT for EMS, and qualitative questions regarding their therapeutic purposes, emotional experiences of using, and perceived helpfulness and rationales. Thematic analysis was used to analyze qualitative data. Results: Most participants reported using ChatGPT for EMS at least 1-2 times per month for purposes spanning traditional mental health needs (diagnosis, treatment, and psychoeducation) and general psychosocial needs (companionship, relational guidance, well-being improvement, and decision-making). Users reported various emotional experiences during and after use for EMS (eg, connected, relieved, curious, embarrassed, or disappointed). Almost all users found it at least somewhat helpful. The rationales for perceived helpfulness include perceived changes after use, emotional support, professionalism, information quality, and free expression, whereas the unhelpful aspects include superficial emotional engagement, limited information quality, and lack of professionalism. Conclusions: Despite the absence of ethical regulations for EMS use, GenAI is becoming an increasingly popular self-help tool for emotional and mental health support. These results highlight the blurring boundary between formal mental health care and informal self-help and underscore the importance of understanding the relational and emotional dynamics of human-GenAI interaction. There is an urgent need to promote AI literacy and ethical awareness among community users and health care providers and to clarify the conditions under which GenAI use for mental health promotes well-being or poses risk.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.2196/77951

Cites

3 of 3 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • LLM-CBs' constant availability, non-judgmental attitudes, knowledge of psychotherapeutic concepts, and perceived empathy make them appear attractive as mental health support tools, especially in societies with logistical and resource-based limitations in mental health treatment.

    “LLM-CBs’ constant availability, non-judgmental attitudes, knowledge of psychotherapeutic concepts, and perceived empathy make them appear as attractive mental health support tools149,150, especially in societies marked by logistical and resource-based limitations in mental health treatment151.”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [149]

    Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • The mental health support potential of LLM-CBs has been noted by both patients and practitioners.

    “The mental health support potential of LLM-CBs has been noted by both patients149,152,153 and practitioners153–155”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [149]

    The Abstract (Methods and Results; no page number available) describes a study of ChatGPT users and reports that almost all found it at least somewhat helpful for emotional and mental health support. The paper recommends that health care providers prepare to discuss AI with clients, but it does not report practitioners’ views of its support potential. Thus it supports the user side of the claim, not the practitioner side. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Risk classification and board-style diagnostic reasoning are poor proxies for the open-ended, ambiguous, multi-turn help-seeking that likely constitutes most real usage.

    “and risk classification and board-style diagnostic reasoning are poor proxies for the open-ended, ambiguous, multiturn help-seeking that likely constitutes most real usage (Raji et al., 2025;Stade et al., 2024;Luo et al., 2025).”

    From HealthBench-Psych: A Mental Health Subset of OpenAI's HealthBench (2026) · cites this paper as Luo et al., 2025

    The Results (Qualitative Results) describe varied help-seeking, including distress reduction, relationship navigation, companionship, and personal decisions; the Discussion (Principal Findings) says a majority of participants sought relief from distress. This supports the claim that real help-seeking extends beyond diagnostic tasks. The paper does not assess risk-classification or board-style tests as proxies, establish that interactions are typically multi-turn, or show that these patterns constitute most real-world usage beyond its sample. (LLM, conf=0.83)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 10

  1. [1]McGrath JJ Al-Hamzawi A Alonso J et al Age of onset and cumulative risk of mental disorders: a cross-national analysis of population surveys from 29 countries Lancet Psychiatry Sep 2023 10 9 668 681 doi 10.1016/S2215-0366(23)00193-1 Medline 37531964 PMC10529120 · DOI · Paper page
  2. [2]Abramson B Boerma J Tsyvinski A Macroeconomics of mental health Natl Bur Econ Res 2024 w32354 doi 10.3386/w32354 · DOI
  3. [3]Gureje O Lasebikan VO Use of mental health services in a developing country. Results from the Nigerian survey of mental health and well-being Soc Psychiatry Psychiatr Epidemiol 01 2006 41 1 44 49 doi 10.1007/s00127-005-0001-7 Medline 16341828 · DOI
  4. [6]Maurya RK Montesinos S Bogomaz M DeDiego AC Assessing the use of ChatGPT as a psychoeducational tool for mental health practice Couns and Psychother Res Mar 2025 25 1 e12759 doi 10.1002/capr.12759 · DOI
  5. [7]Arjanto P Senduk FFW Literature review on the double-edged sword of ai in mental health: a deep dive into ChatGPT’s capabilities and limitations CMHP 2024 6 2 67 76 doi 10.51602/cmhp.v6i2.144 · DOI
  6. [8]Luo X Ghosh S Tilley JL Besada P Wang J Xiang Y “Shaping ChatGPT into my Digital Therapist”: a thematic analysis of social media discourse on using generative artificial intelligence for mental health Digit Health 2025 11 20552076251351088 doi 10.1177/20552076251351088 Medline 40656852 PMC12254646 · DOI
  7. [9]Siddals S Torous J Coxon A “It happened to be the perfect thing”: experiences of generative AI chatbots for mental health Npj Ment Health Res Oct 27 2024 3 1 48 doi 10.1038/s44184-024-00097-4 Medline 39465310 PMC11514308 · DOI · Paper page
  8. [11]Douglas BD Ewell PJ Brauer M Data quality in online human-subjects research: comparisons between MTurk, Prolific, CloudResearch, Qualtrics, and SONA PLOS ONE 2023 18 3 e0279720 doi 10.1371/journal.pone.0279720 Medline 36917576 PMC10013894 · DOI · Paper page
  9. [13]Levitt HM Bamberg M Creswell JW Frost DM Josselson R Suárez-Orozco C Journal article reporting standards for qualitative primary, qualitative meta-analytic, and mixed methods research in psychology: the APA Publications and Communications Board task force report Am Psychol 01 2018 73 1 26 46 doi 10.1037/amp0000151 Medline 29345485 · DOI · Paper page
  10. [14]Cohen KA Stiles-Shields C Winquist N Lattie EG Traditional and nontraditional mental healthcare services: usage and preferences among adolescents and younger adults J Behav Health Serv Res Oct 2021 48 4 537 553 doi 10.1007/s11414-020-09746-w Medline 33474642 PMC7817073 · DOI

Cited by

Sorted by citations

  • A scoping review on the mental health harms of LLM-based chatbots

    Diel, A., Torous, J., Cuijpers, P., Kleesiek, J., Nensa, F., Weber, N., Faust, F., Lalgi, T. J., Mellis, F. S., Teufel, M., et al. (2026) · npj Digital Medicine · doi:10.1038/s41746-026-03054-x · Cited by 4 · see citing works

    “The mental health support potential of LLM-CBs has been noted by both patients 149,152,153 and practitioners 153 – 155 , and it is expected that the role of LLM-CBs in mental health care will continue to grow 156 .”

  • HealthBench-Psych: A Mental Health Subset of OpenAI's HealthBench

    Flathers, M., Nguyen, P. A., Noorily, J., Herpertz, J., Chen, M., Multani, J., Powell, S., Granof, M., Kalinch, M., & Torous, J. (2026) · arXiv · doi:10.48550/arxiv.2608.25071

    “…detection and diagnostic classification (Flathers et al., 2026), and risk classification and board-style diagnostic reasoning are poor proxies for the open-ended, ambiguous, multi-turn help-seeking that likely constitutes most real usage (Raji et al., 2025; Stade et al., 2024; Luo et al., 2025).”

Authors
Luo, X., Wang, Z., Tilley, J. L., Balarajan, S., Bassey, U.-A., & Cheang, C. I.
Affiliations
  • Santa Clara University (US)
  • Nanyang Technological University (SG)
Corresponding author
Luo, Xiaochen
ORCID
Venue
JMIR Mental Health
Publisher
JMIR Publications Inc.
Published
2025-11-27
Volume
12
Pages
e77951-e77951
ISSN
2368-7959 (ISSN-L)
DOI
10.2196/77951
PMC
PMC12661908
OpenAlex
W4414967761
Cited by
48
References
3
Citations by year
2026: 48
FWCI
38.41
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • JMIR Mental Health (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • DR-NTU (Nanyang Technological University) (submittedVersion)
  • NIE Digital Repository (National Institute of Education, Nanyang Technological University) (submittedVersion)
Topics
  • Social Sciences › Psychology › Applied Psychology › Digital Mental Health Interventions
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
Keywords
ChatGPT, generative AI, mental health support, emotional support, user experience, perceived helpfulness, self-help, human-AI interaction, AI literacy, AI ethics, psychoeducation, companionship, well-being
MeSH terms
Generative Artificial Intelligence, Adult, Artificial Intelligence, Mental Disorders, Emotions, Female, Humans, Male, Mental Health, Mental Health Services, Middle Aged, Surveys and Questionnaires, Qualitative Research, Young Adult

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review