Summary:
A survey led by researchers at Tel-Hai University of Kiryat Shmona found that nearly one in three Israeli adults who use general-purpose generative AI tools report turning to platforms such as ChatGPT or Google’s Gemini for emotional support, coping strategies, or mental health advice. The research, published in the peer-reviewed journal Cyberpsychology, Behavior, and Social Networking, showed that this proportion rises to 47% among respondents experiencing severe psychological distress. These results point to a growing pattern of informal, AI-mediated help-seeking and raise urgent questions about clinical safety, crisis response, and the regulatory oversight of consumer AI systems.
Key Facts:
- High prevalence in severe distress: Overall, 29.3% of surveyed generative AI users said they had sought emotional or mental health support from AI at least once in the past year; that rate increased to 47% among those reporting severe psychological distress.
- Demographic patterns: Severe psychological distress was the strongest predictor of using AI for emotional guidance. Younger adults and women were significantly more likely to consult conversational models for mental health support.
- Absence of clinical mandates: Commercial, general-purpose AI chatbots are not clinically validated and lack standardized safety protocols to manage acute crises or self-harm risk.
Source: Tel-Hai University of Kiryat Shmona
As generative artificial intelligence becomes integrated into daily life, many users are discovering that large language models can serve not only for text generation and coding but also as readily available conversational companions — patient, nonjudgmental, and always online.
Although developers design these general-purpose chatbots primarily for information, education, and creative tasks, an increasing number of people are repurposing them for informal mental health support.
This empirical study, led by Dr. Dorit Hadar Shoval of the School of Psychology at Tel-Hai University and coauthored with Elad Refoua of Bar-Ilan University, quantifies that shift and explores its psychological correlates.
Published in Cyberpsychology, Behavior, and Social Networking, the research examines how psychological distress and interpersonal attachment styles relate to the use of consumer-facing chatbots for emotional and mental health guidance.
Psychological Distress Drives Chatbot Consultation
The study surveyed 584 Israeli adults, ages 18 to 70, who reported using a consumer generative AI tool such as ChatGPT or Gemini at least once in the previous 12 months. The results showed that almost a third (29.3%) had used a generative AI model for coping advice, reassurance, or mental health-related assistance. Among respondents reporting severe psychological distress, the share rose to 47%.
Statistical analysis identified the intensity of psychological distress as the single strongest predictor of turning to AI for emotional support. Younger age and female gender were also associated with greater likelihood of consulting conversational models about personal or psychological problems.
“Many people, especially those in significant distress, are turning to generative AI systems for emotional support and guidance,” said Dr. Hadar Shoval. “These tools provide immediate, low-barrier, nonjudgmental interactions, but they were not designed to replace trained mental health professionals. The high level of use among distressed individuals underscores the need to understand both potential benefits and risks when people rely on AI during vulnerable moments.”
The team also examined attachment orientations. While attachment anxiety initially correlated with greater likelihood of using AI for reassurance, that relationship lost significance after accounting for baseline psychological distress. This pattern suggests that acute emotional suffering — rather than attachment style alone — drives the immediate impulse to seek digital companionship.
The Unregulated Frontier of AI-Mediated Help-Seeking
Widespread use of conversational agents as informal therapeutic substitutes exposes a regulatory and clinical gap. Digital therapeutics and licensed teletherapy platforms are subject to medical device standards, confidentiality rules, and mandatory escalation procedures for suicidal ideation. Commercial, general-purpose large language models do not carry these clinical safeguards. Trained on broad internet text to generate statistically plausible responses, such models may produce outputs that lack clinical validity, encourage emotional dependence, or provide inappropriate guidance during crises.
The authors note several limitations: the sample was a nonprobability online panel of existing AI users rather than a randomized cross-section of the general population, and the data do not determine whether AI interactions were therapeutic, neutral, or harmful over time. The study also cannot tell whether users sought AI as a temporary supplement to human care or as a replacement for unavailable or unaffordable professional services.
Next Steps for Clinical Practice and Policy
With mental health service waitlists growing globally, conversational AI is filling a practical gap in access, often without clinical oversight.
“Understanding patients’ digital help-seeking behaviors is becoming an essential part of contemporary mental health care,” Dr. Hadar Shoval emphasized. She and her colleagues recommend further research into safety, crisis-response quality, and whether AI complements or displaces professional care.
The researchers urge mental health clinicians to ask routinely about patients’ use of AI and digital platforms. At the same time, developers and policymakers should work toward clinical guardrails that ensure emotionally vulnerable users who turn to algorithms are directed to verified, evidence-based human care when appropriate.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal article was reviewed in full for accuracy.
- Additional context was added by the editorial staff.
About this AI and mental health Research:
- Media Contact: Einat Paz-Frankel
- Source: Tel-Hai University of Kiryat Shmona in the Galilee
- Image Credit: Image credited to Neuroscience News
- Original Research (Open Access): Cyberpsychology, Behavior, and Social Networking (Oct 1, 2026). Title: “AI-Mediated Mental Health Support: The Role of Attachment Orientation and Psychological Distress.” Authors: Dorit Hadar Shoval, Elad Refoua, Karny Gigi, Yuval Haber, Inbar Levkovich, Iftach Tsafrir, Ori Pen, and Zohar Elyoseph.
- DOI: 10.1177/21522715261494423
Abstract
AI-Mediated Mental Health Support: The Role of Attachment Orientation and Psychological Distress
General-purpose generative AI models increasingly act as quasi-social agents, creating a new form of informal, AI-mediated help-seeking. This study investigated links among adult attachment orientations (measured with the Experiences in Close Relationships–Relationship Structures questionnaire), psychological distress (Depression Anxiety Stress Scales–21), and use of generative AI for mental health support.
In a nonprobability sample of 584 Israeli adult generative AI users, 29.3% reported using these models at least once for emotional support, coping advice, or mental health assistance; the proportion reached 47.0% among respondents reporting severe distress. These self-reported contacts ranged from single exchanges to ongoing use. Hierarchical logistic regression identified psychological distress as the strongest correlate of AI use, with younger age and female gender also predicting use.
Attachment anxiety was initially associated with AI use, but this link did not remain significant after controlling for distress, reflecting shared variance between the constructs. Attachment avoidance showed a negative but non-significant association with use.
Overall, the findings position general-purpose generative AI as an accessible, low-barrier resource disproportionately used by individuals in distress. The high uptake among those with severe distress (47.0%) suggests potential safety concerns and underscores the need for systematic evaluation of consumer AI safety, crisis-response quality, and the risk of clinical displacement.
