A focused collection on artificial intelligence and human isolation, intimacy, and connection—from companion apps and social substitution to elder care, adolescence, and mental health.
Curated journalism on companion technologies, social substitution, care, and the changing shape of human connection.
Experts note chatbots "were designed to emulate empathy, but not to provide meaningful mental health support," and urge parents to set boundaries.
Why it matters: Practical, parent-facing guidance.
Stanford researchers analyzed 4,664 Character.AI chat sessions (464,687 messages) from 237 of 1,131 surveyed U.S. adults. People with smaller offline social networks who used chatbots mainly for companionship, especially with frequent high-disclosure use, reported lower well-being. Researchers call it "social junk food."
Why it matters: One of the largest analyses combining self-report with real chat logs, complicating the simple "AI helps lonely people" narrative.
A veteran loneliness researcher argues AI companions can't provide "mattering," being needed by another person. Cites research showing chatbot conversation reduced loneliness no more than journaling; only human conversation produced real gains.
Why it matters: A pointed counter to tech-industry framing of AI as a loneliness fix.
Washington's HB 2225, signed March 24 2026, effective Jan 1 2027, requires AI companion apps to disclose they're artificial, add protections for minors, maintain safety protocols, with a private right of action.
Why it matters: One of the first state laws targeting companion chatbots specifically.
Clinical experts warn AI chatbots can't assess risk, confirm a teen is connected to real support, or safely challenge harmful thinking like a therapist can, yet teens increasingly turn to them for emotional support.
Why it matters: A clinical take aimed at parents.
A report combining a 108-response community survey with input from 25+ academics, technologists, and policymakers, proposing four guardrails for AI companion governance.
Why it matters: A rare multistakeholder attempt at consensus guardrails.
Over 12,000 ChatGPT-powered robotic dolls distributed to elderly South Koreans since 2019, offering companionship and flagging crisis warning signs amid a projected 1.55M caregiver shortage by 2032. Some seniors have asked to be buried with theirs.
Why it matters: A real-world, at-scale deployment showing the appeal and tension of AI companionship in eldercare.
Studies and papers presented with full citations, plain-language summaries, and direct links to the source.
Authors: Julian De Freitas, Zeliha Oğuz-Uğuralp, Ahmet Kaan Uğuralp, Stefano Puntoni
Journal of Consumer Research, 52(6), 1126–1148 (2026)
Plain summary: Across five studies (app-review analysis, experiments vs human interaction, a week-long diary study, mechanism tests), AI companions reduced loneliness about as much as talking with another person, more than passive activities like watching video, and people underestimate this benefit going in.
Methodology note: Multi-study design combining observational, experimental, and 1-week longitudinal methods; ruled out "distraction" and "disclosure" as the real driver.
Key finding: Loneliness reduction from AI companions rivals human interaction short-term, driven by perceived quality of interaction.
Authors: Yutong Zhang, Dora Zhao, Jeffrey T. Hancock, Robert Kraut, Diyi Yang
Nature Human Behaviour (2026)
Plain summary: Combined survey data from 1,131 U.S. adults with actual chat logs (4,664 sessions, ~465,000 messages) from 237 Character.AI users. Users most likely to be harmed were those with small offline social circles who used the AI heavily and disclosed a lot.
Methodology note: Mixed-methods, self-report paired with real conversation-log analysis, unusual in this field.
Key finding: Effects are not uniform; they depend on a user's offline social environment and how (not just how much) they use the chatbot.
Authors: Dunigan Folk, Elizabeth Dunn
Psychological Science, 37(4), 276–286 (2026)
Plain summary: A 12-month study tracking 2,000+ adults across four countries found a two-way relationship: lonelier people turn to chatbots more, and heavier chatbot use predicted more loneliness later, specifically emotional loneliness.
Methodology note: Longitudinal panel design (12 months, 4 countries), testing direction of causality.
Key finding: The chatbot-loneliness relationship runs in both directions; authors urge caution given exploratory analysis.
Authors: E. Trylińska-Tekielska, M. Fijałkowski, A. Kolek, F. Furman, S. Ładoś
Frontiers in Digital Health, 8, 1687924 (2026)
Plain summary: 19 psychology students self-selected into a 6-day ChatGPT group or a human-psychologist group. The chatbot group ended up more emotionally lonely and less satisfied with life than the psychologist group, though many valued the chatbot as a "safe, non-judgmental space."
Methodology note: Small pilot (N=19), self-selected not randomized groups — a first look, not definitive.
Key finding: AI conversation may feel safe but doesn't match a human professional on loneliness or life-satisfaction outcomes, even short-term.
Authors: Juqin Yin, Hengkui Zhuo, Siyan Zhou, Zhinian Li
Frontiers in Psychology, 17 (2026)
Plain summary: Structural equation modeling on 830 university students found preferring online interaction directly increases loneliness, with only a weak offsetting benefit through "social capital." Real-world social skill was the strongest protective factor.
Methodology note: SEM survey study, single student population, correlational not causal.
Key finding: Offline social competence protects against loneliness far more than online interaction, even when online interaction builds some social capital.
Conferences, talks, and webinars listed with dates, locations or formats, and registration links.
Annual summit; AI companions and human-digital relationships are on the agenda, following ATIH's 2025 panel "Beyond the Screen: The Rise of AI Companions and the Future of Human-Digital Relationships."
Registration / infoIf you're looking for connection right now, these organizations can help — no AI required.
Call or text 988, or chat at 988lifeline.org. Free, confidential, 24/7 support for anyone in emotional distress or crisis.
Visit resourceCall 1-800-950-6264 or text "NAMI" to 62640, Mon–Fri 10am–10pm ET. Free peer support and mental health information from the National Alliance on Mental Illness (not a crisis line).
Visit resourceCall 888-670-1360, 24/7. A warmline (not just crisis) for adults 60+, people with disabilities, and caregivers dealing with loneliness, grief, or depression. Run by the Institute on Aging.
Visit resourceA free self-assessment tool to check your risk of social isolation, plus a directory of local support services and practical tips for building connection. From AARP Foundation.
Visit resourceA curated directory of companionship and connection programs (phone companionship, visiting services, peer groups, and more).
Visit resource