Interaction with AI companions and psychological well-being
Stanford-led study of 1,131 adult Character.AI users combining survey self-report with donated chat transcripts from 244 of them, analyzed with LLM-assisted methods against the Comprehensive Inventory of Thriving. Examines how usage motivation, interaction intensity, self-disclosure, and offline social context relate to psychological well-being, finding that companionship-motivated use among people with smaller real-world social networks is associated with lower well-being.
Publisher
Nature Human Behaviour
Published
4 Aug 2026
Added
2 months ago
Key Findings
- Just under 12% of surveyed users reported companionship as their primary motive, yet over 50% described the chatbot in relational terms (friend, companion, romantic partner) and more than 80% of donated chat sessions centered on seeking emotional or social support
- Intense chatbot use among participants with smaller real-world social networks correlated with poorer well-being, with the association strongest when companionship motivated the use
- Greater willingness to disclose sensitive personal information to the AI companion was associated with lower well-being — the opposite of the pattern established for self-disclosure in human relationships
Methodology Notes
Survey of 1,131 US adult Character.AI users plus donated chat transcripts from 244 participants; transcripts analyzed with GPT-4o, LLaMA-3-70B, and TopicGPT; well-being measured with the Comprehensive Inventory of Thriving. Published online 2026-08-04. The nature.com landing page auth-walls automated clients; verified via Crossref DOI metadata (title, authors, journal, online date) and Stanford HAI's article on the study.
Sources
Nature Human Behaviour article page(opens in a new tab) (primary)
Stanford HAI coverage(opens in a new tab) (5 Aug 2026)
Topics
Authors
Yutong Zhang, Dora Zhao, Jeffrey T. Hancock, Robert Kraut, Diyi Yang
Tags
Cite This
APA
Yutong Zhang et al. (2026). Interaction with AI companions and psychological well-being. Nature Human Behaviour. https://www.nature.com/articles/s41562-026-02516-2
Related Insights
AI companions and subjective well-being: Moderation by social connectedness and loneliness
Technology in Society (Elsevier) · 1 Apr 2026
Investigating Affective Use and Emotional Well-being on ChatGPT
OpenAI; MIT Media Lab · 4 Apr 2025
Why human–AI relationships need socioaffective alignment
Humanities and Social Sciences Communications (Nature Portfolio) · 28 May 2025
Beyond Benefit or Risk: Perceived Impact Profiles of Human-AI Affective Interaction and Their Associations with Psychological Functioning
arXiv (Beijing Normal University, Faculty of Psychology; The Ohio State University) · 15 Sept 2026
AI emotional support is better only when chosen, but shifts preferences even when it is not
arXiv (preprint) · 24 Aug 2026
Longitudinal Evidence That General-Purpose Chatbots Actively Foster Relational Engagement
arXiv preprint · 11 Aug 2026
Friends, Lovers, and Companions: A Systematic Scoping Review of Empirical Evidence on Human-AI Relationships
Computers in Human Behavior: Artificial Humans (Elsevier) · 1 Aug 2026
The Rise of AI Companions: A national public opinion survey on the relationships people form with artificial intelligence systems
Elon University Imagining the Digital Future Center · 15 Jun 2026
Adolescent AI Use and Evaluations: Prospective Bidirectional Associations with Internalizing Symptoms
PsyArXiv (University of North Carolina at Chapel Hill) · 3 Aug 2026
Randomized Trial of AI-Powered Mental Health Support Embedded in University Academic Infrastructure
PsyArXiv; Stanford University; University of Manitoba; Flourish Science; Harvard Business School; University of California, Berkeley; Massachusetts General Hospital / Harvard Medical School · 10 Sept 2026
How Does Turning to AI for Companionship Predict Loneliness and Vice Versa?
Psychological Science (Association for Psychological Science / Sage) · 23 Mar 2026