Skip to main content

Issue #005 — JAMA Pediatrics puts a peer-reviewed number on the demand side: ~20% of US youth now use AI chatbots for mental-health advice, and 63% hide it.

Isuru Gunarathne
Software engineer & researcher

Weekly Intelligence · Week 5 · 5 June 2026 · Issue #005​

A new JAMA Pediatrics survey puts a peer-reviewed number on the demand side: ~20% of US 12–21-year-olds used AI chatbots for mental-health advice in 2025, up from 13% a year earlier, and 63% disclosed it to no one.

Executive Summary​

This was a quiet week for primary detection research and a sharp one for measuring the demand side. The single firmly-dated, peer-reviewed development is McBain et al., AI Chatbot Use and Disclosure for Mental Health Among US Adolescents and Young Adults, published in JAMA Pediatrics on 3 June 2026 — a nationally-weighted survey (1,009 respondents aged 12–21, fielded November 2025, weighted to ≈42 million US youth) finding that 19.2% had used AI chatbots for mental-health advice in 2025, up from 13.1% the prior year, with a striking 63.3% disclosing that use to no one. The study converts threads this newsletter has tracked qualitatively and via single-session conference figures (the Drexel "bond paradox," Issue #004; the JHU "AI for Hope" 5-million-young-people estimate, Issue #004; the 16%-of-adults figure, Issue #001) into a peer-reviewed, demographically-stratified denominator — and adds a genuinely new variable, non-disclosure, that detection and clinical-intake pipelines have not previously been able to size. No new wearable, speech, or multimodal primary results cleared the 7-day window; several candidate papers surfaced in search but date to February–April 2026 (e.g. a Journal of Affective Disorders acoustic-biomarker analysis) or were already covered (the Mindcraft adolescent digital-phenotyping feasibility study, Issue #001), and are held out. The honest read: the field's measurement frontier was static this week, but the best evidence yet on how many young people are using these tools, how often, and how secretly landed in a top journal.

Key Metrics​

MetricValueSource
US youth (12–21) using AI chatbots for mental-health advice, 2025 vs 202419.2% vs 13.1%McBain et al. · JAMA Pediatrics · 3 Jun 2026
Users who disclosed the chatbot use to no one63.3%McBain et al. · JAMA Pediatrics · 3 Jun 2026
Users finding the experience helpful / using ≥ monthly91.7% / >40%McBain et al. · JAMA Pediatrics · 3 Jun 2026

Clinical Translation & Help-Seeking Behavior​

JAMA Pediatrics: ~20% of US youth use AI chatbots for mental-health advice — and 63% hide it​

A team led by Ryan K. McBain (RAND), with collaborators at Harvard Medical School and the MIT Media Lab, surveyed a nationally representative sample of 1,009 US adolescents and young adults aged 12–21 in November 2025, statistically weighted to represent roughly 42 million youth. The headline: 19.2% reported using AI chatbots for mental-health advice during 2025 — an estimated 8.2 million young people — up sharply from 13.1% in the equivalent 2024 survey. Of those users, 63.3% had not disclosed the practice to anyone; more than 40% used a chatbot for such advice at least monthly and 5.8% did so daily or almost daily; and 91.7% rated the experience as helpful. Use was higher among girls and young women, among older teens, and among those who had recently consulted a physician about mental health. The authors caution that the high perceived helpfulness may reflect chatbots' tendency to be overly agreeable or flattering rather than the accuracy of the advice — the same sycophancy-and-reassurance failure mode the Mpathic benchmark (Issue #002) and the Drexel "bond paradox" (Issue #004) characterised from the model and relational-behavior sides respectively. For this newsletter the new and load-bearing variable is non-disclosure: a near-two-thirds hidden-use rate means clinician-intake screening for chatbot use (the WBUR / JAMA Psychiatry intake practice flagged in Issue #001) will systematically under-count exposure unless it is actively and non-judgmentally prompted, and it sets a hard ceiling on how much of this behavior any passively-observed clinical signal can currently see.

Source: McBain RK, et al. · JAMA Pediatrics · 3 Jun 2026 · 10.1001/jamapediatrics.2026.2015 Source: Medical Xpress · "1 in 5 teens turn to AI chatbots for mental health advice…" · 3 Jun 2026 · medicalxpress.com Source: American Journal of Managed Care · "AI Chatbot Use for Mental Health Advice Rises Sharply Among US Youth" · Jun 2026 · ajmc.com

Forward Outlook​

  • Near-term: The 63% non-disclosure figure is the citable number that turns "ask patients about chatbot use" from a recommendation into a screening-design requirement. Expect it to appear quickly in intake-questionnaire guidance and in the state-legislative testimony tracked since Issue #001 (Utah HB 452) and Issue #004 (the 800-bill / 3-enacted review) — hidden use by minors is the most legible argument for disclosure mandates.
  • Mid-term: Paired with the GBD 2023 adolescent-peak shift (Issue #003), the McBain prevalence trajectory (13% → 19% in one year) strengthens the case for the triage-grade, human-in-the-loop deployment shape this newsletter has argued toward — and specifically for routing-into-care designs over autonomous companionship, given that the heaviest, most hidden use sits in exactly the cohort with the fastest-rising burden.
  • Long-term: Non-disclosure is now a measured ceiling on observational detection. If most at-risk chatbot use is invisible to clinicians and to passive sensing alike, the field's early-detection value increasingly depends on instrumentation inside the chatbot surface (guardrails, bond-paradox-aware routing, VERA-MH / Mpathic-style safety evaluation) rather than on external behavioral signals catching the behavior after the fact.

Sources used: 3 · Week 5 · Next issue: 12 June 2026