The first randomized trial of a generative-ai therapy chatbot found that 210 adults rated their…
the first randomized trial of a generative-ai therapy chatbot found that 210 adults rated their alliance with the machine as comparable to human psychotherapy, and symptom gains exceeded what is commonly reported for antidepressants. openai followed by shipping mentalhealthbench, an expert-informed evaluation set for ai crisis handling.
the shortage of human therapists is becoming the argument for the chatbot, not against it.
Context
The NEJM AI paper 'Randomized Trial of a Generative AI Chatbot for Mental Health Treatment' (AIoa2400802) is the trial of the Therabot chatbot. EurekAlert (Mar 27, 2025) headlined it 'First therapy chatbot trial shows AI can provide gold-standard care'.
OpenAI introduced MentalHealthBench on Sep 23, 2026: an open benchmark developed with more than 80 licensed mental health experts, with 1,215 synthetic mental health conversations and weighted rubric criteria (OpenAI; nope.net).
The first randomized trial of a generative AI therapy chatbot was published and reported in March 2025, so it is not new, and OpenAI's benchmark followed it by about 18 months rather than immediately. The 210 adults, the alliance comparable to human psychotherapy and symptom gains exceeding antidepressant effects were not seen in the sources read, so unsupported here, not refuted. in the excerpts read. MentalHealthBench is an expert-informed evaluation of AI responses in realistic mental health conversations. 'The shortage of therapists is the argument for the chatbot' is the author's opinion.
Related work
- OpenAI: MentalHealthBench, an expert-informed benchmark ↗Benchmark paper.
- nope.net: 2026 OpenAI MentalHealthBench ↗Sep 23, 2026.
- Evaluating Therabot: a randomized control trial ↗Jun 14, 2024.
Watch next
- Find the participant count and effect sizes in the NEJM AI paper.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 1 October 2026 at 18:17 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →