Mental health AI
Therabot
Therabot is a generative AI therapy chatbot developed since 2019 at Dartmouth and tested in the first randomized controlled trial of a generative AI chatbot for mental health treatment, published in NEJM AI in March 2025, which reported a 51 percent average reduction in depression symptoms, 31 percent in anxiety, and 19 percent in eating-disorder concerns against a waitlist control, with human clinicians monitoring conversations. As of this review it remains research-only: no public product was found, a Therabot-CALM trial for cannabis use with anxiety or depression is recruiting, and the inventors have co-founded Therabot Labs, LLC to license the technology. Its CAIHL significance is the evidence model itself: clinician-built, trial-gated, safety-monitored generative therapy, which is the standard the marketed consumer chatbots have not met.
This profile is using an explicit publication status from the profile data.
Summary judgment · 45% toward patient-directed
Research-stage, unclear in deployment
Therabot is patient-facing in use but researcher-controlled in access: participants choose to enroll, yet the public cannot choose it at all, and its eventual posture depends on whether the trial-grade safeguards survive commercialization through Therabot Labs, LLC.
Patient agency
How this tool changes agency
The RCT reported clinically significant symptom reductions and therapeutic alliance ratings comparable to human therapists, and the app prompted users to call 911 or crisis lines when high-risk content appeared.
Participation is voluntary with consent and withdrawal rights, but the public cannot choose Therabot at all because access is gated by study enrollment.
Patient-facing signals
Who does this AI serve?
The tool currently serves supervised trial participants and the scientific case for safe generative therapy, with no advertising or data monetization, but the study authors' spinout company introduces future commercial interests that should be tracked.
Can patients tell AI is involved?
Participants enroll through informed consent knowing Therabot is an AI chatbot, and the trial design disclosed AI involvement and human monitoring.
Can patients meaningfully choose?
Participation is voluntary with consent and withdrawal rights, but the public cannot choose Therabot at all because access is gated by study enrollment.
Can patients correct or challenge what the AI produces?
Trial conversations were reviewed by the study team, which was equipped to intervene if the software responded outside best practices, and IRB oversight applies, but no patient-facing correction workflow exists because there is no public product.
Does it help patients understand or act?
The RCT reported clinically significant symptom reductions and therapeutic alliance ratings comparable to human therapists, and the app prompted users to call 911 or crisis lines when high-risk content appeared.
Text findings
Who is left out or burdened?
Everyone outside the trials, by design
Trial gating excludes the general public, which is honest but limits reach; trial access required a smartphone, English, and US residence, and the equity profile of any future product will depend on licensing and pricing decisions not yet disclosed.
What happens to patient data?
Research-consent governed; future product practices not disclosed
Trial data handling falls under informed consent and IRB oversight, with researchers reviewing conversations for safety; data practices for any commercial version through Therabot Labs, LLC are not disclosed because no product exists.
Are the clinical boundaries clear?
Clear in the trial design
High-risk content triggered prompts to call 911 or crisis lines, the study team monitored and could intervene, and the authors themselves state that no generative AI agent is ready to operate fully autonomously in mental health.
Who defined what good looks like?
Clinician and researcher defined, peer reviewed
Outcomes used standardized clinical measures in a registered randomized trial published in NEJM AI and funded by Dartmouth, the strongest evaluation ownership in this category, with the caveat that the evaluators are also the founders of the commercializing spinout.
Review method
Public-source review of the NEJM AI trial publication record, Dartmouth press releases from 2025 and 2026, ClinicalTrials.gov registration, and trial-listing pages; no contact with the research team, no app access, and no review of unpublished follow-up data.
Research stage, not publicly available · Medium draft, peer-reviewed RCT plus institutional sources, commercialization timeline and access status partly unclear