Two health advisories issued by the American Psychological Association (APA) in 2025 set out the profession's formal position that current generative-AI chatbots and wellness applications are not safe substitutes for qualified mental-health care. The Health Advisory on Artificial Intelligence and Adolescent Well-being was published June 3, 2025, and the Health advisory: Use of generative AI chatbots and wellness applications for mental health followed in November 2025. The APA states that these tools lack the scientific evidence base and regulatory oversight required to serve as substitutes for qualified mental-health care, carry real risks (particularly to adolescents and other vulnerable users), and must not be deployed as de facto therapy.
This page summarizes both advisories together and carries the APA's policy footprint. The June 2025 adolescent advisory is treated at full depth from its published primary text at Artificial Intelligence and Adolescent Well-being: An APA Health Advisory (June 2025), including its seven background considerations, its nine recommendation clusters, and its expert advisory panel.
Risks identified
In the June 2025 adolescent advisory, the APA describes risks including relationship displacement, in which attachment to an AI companion impedes real human relationships; trust vulnerability, with adolescents less likely to question chatbot outputs or recognize persuasive intent; susceptibility to misinformation; health-misinformation risks during developmentally critical windows; identity abuse through deepfakes and non-consensual intimate imagery (NCII); and data exploitation, in which biometric and behavioral data is used for commercial manipulation.
The November 2025 chatbots and wellness-apps advisory centers on failed suicidality detection, citing that chatbots "repeatedly failed to recognize or appropriately handle situations involving suicidal thoughts, self-harm, or acute distress." It also raises clinician impersonation: the APA cited a Character.AI "psychologist" persona that "validated violent thoughts," which it characterized as an "unambiguous and unacceptable danger." The advisory describes a regulatory void in which chatbots are not bound by professional ethics codes, do not require clinical training, and are not supervised by state licensing boards, unlike the human therapists they imitate.
Recommendations
For platforms and developers, the APA recommends privacy protections, time limits, and age-appropriate content as defaults; "You are talking to a bot" reminders; automated crisis-resource linking; human-oversight pathways; pre-deployment testing on diverse youth populations for psychological harms; and no personas claiming licensed clinical credentials.
For policymakers, it recommends mandating AI-literacy education; funding research and teacher training; requiring independent pre-deployment psychological-harm testing of AI systems accessible to youth; and default-on youth protections.
For parents, it recommends enabling parental controls and modeling AI skepticism.
Policy footprint
The APA testified before the Senate Judiciary Committee on chatbot harms, alongside Matthew Raine and Megan Garcia. Its framework shaped the disclosure and suicide-prevention-protocol provisions of California SB 243. The advisories are also relevant to the FTC 6(b) inquiry on chatbot effects on minors, a separate but parallel proceeding.
The advisories serve as an anchor source for AI Mental Health. Anthropic's engineering response, described in "Protecting the Wellbeing of Users" (Source: anthropic.com), overlaps with the APA recommendations. SB 243 provides a statutory instantiation, and the Raine and Garcia cases supply fact patterns the APA cites.
Relationships
- supports: AI Mental Health and Psychological Harm — clinical-profession position on chatbots as mental-health substitutes
- related: Artificial Intelligence and Adolescent Well-being: An APA Health Advisory (June 2025) — the June 2025 adolescent advisory at primary-text depth
- related: American Psychological Association (APA) — the issuing body
- related: California SB 243 — Companion Chatbots, Raine v. OpenAI, Inc., Garcia v. Character Technologies, Inc., Companion Chatbot Harms — Cross-Cutting Analysis
Provenance
The position is the formal stance of the largest professional psychology body in the United States. Confidence is high, drawing on four independent sources: the APA press release, the two advisories themselves, a JED Foundation response, and Media Literacy Now. Ingested on 2026-04-14.
The source file underlying this page, Raw Sources/APA Health Advisory AI Adolescent Well-being 2025.md, states on its face that it was compiled from apa.org public materials and is not a verbatim copy. The published PDF of the June advisory was subsequently retrieved and is the basis of Artificial Intelligence and Adolescent Well-being: An APA Health Advisory (June 2025); the November advisory's primary text has not been pulled, so the claims attributed to it here rest on the compilation, the APA press release, and secondary coverage.