US Offices
14 Wall Street, 20th Floor
10005 NY
United States,
[email protected]
European Offices
Chemin du Vernay 14a,
1196 Gland, CH-Vaud,
Switzerland,
[email protected]
Back

The AI Therapy Paradox: Proof Arrives as the Pioneers Bow Out

TLDR: Spring 2025 delivered a striking split screen in digital mental health. The first randomized trial of a generative artificial intelligence (AI) therapy chatbot reported clinical results that rival traditional care, while one of the field’s pioneers retired its app and pointed to the cost of meeting medical regulation. Capability and approval moved on different clocks, and the gap between them now defines who reaches patients.

The evidence finally arrived

For years, therapy chatbots promised more than the data could support. That changed when researchers at Dartmouth published the first randomized controlled trial of a generative AI therapy chatbot, Therabot, in the journal NEJM AI. The study enrolled 210 participants with major depression, generalized anxiety, or high-risk eating-disorder symptoms.

The results stood out. Participants saw a 51% reduction in depression symptoms, a 31% reduction in anxiety, and a 19% improvement in eating-disorder concerns, with many describing a genuine therapeutic alliance with the software. For the first time, a rigorous trial placed AI therapy in the same conversation as established outpatient care.

A pioneer chose to exit

In the same season, one of the category’s founders stepped back. Woebot Health announced it would retire its app on June 30, 2025, closing a product that roughly 1.5 million people had used over the years. Woebot built its experience on carefully pre-scripted, clinically grounded conversations.

Founder Alison Darcy explained the decision in plain terms: the cost and difficulty of meeting the Food and Drug Administration’s requirements for marketing authorization, combined with the arrival of a powerful new wave of conversational AI that Woebot had foreshadowed. A careful, science-first pioneer chose to close the chapter precisely as the science turned in its favour.

The paradox sits in the timing

Place the two events side by side and the tension becomes clear. Clinical proof for generative AI therapy arrived at the exact moment a leading scripted-bot company wound down. The capability curve and the approval curve crossed in opposite directions.

The explanation lies in how each product was built. Woebot held itself to a medical-device standard, which carries a heavy and expensive regulatory burden. The new generation of generative systems delivers fluid, adaptive conversation, and the rules that govern safety, evidence, and liability for that technology are still taking shape. Capability advanced faster than the framework around it.

Why this matters beyond mental health

The pattern reaches across regulated healthcare. A technology can demonstrate strong outcomes in a trial and still face a long road to routine, reimbursed use. The winners will pair clinical evidence with a credible regulatory and safety strategy, because patients and payers reward proof that holds up under scrutiny.

The same boundary now runs through home care, where the wording of an AI briefing decides whether it is documentation software or a regulated medical device.

Mental health makes the stakes vivid. Demand far outstrips the supply of human clinicians, and accessible, effective, around-the-clock support could reach millions who wait months for care today. Realising that promise depends on building trust into the product: clinical validation, clear guardrails, human oversight where it counts, and transparent handling of sensitive data.

The work ahead

The Therabot result and the Woebot exit together mark a turning point. Generative AI in mental health now has its first strong evidence and its first hard lesson about the distance between a working product and an approved one. The opportunity belongs to teams that treat regulation and safety as design problems to solve early.

That challenge, where powerful technology meets the demands of a regulated industry, is the terrain Kainjoo works in every day. The brands that pair real evidence with real trust will carry AI mental health from a landmark trial into everyday care.

References

  1. First therapy chatbot trial yields mental health benefits (Dartmouth, Therabot, NEJM AI): https://home.dartmouth.edu/news/2025/03/first-therapy-chatbot-trial-yields-mental-health-benefits
  2. Woebot Health shuts down its pioneering therapy chatbot: https://www.statnews.com/2025/07/02/woebot-therapy-chatbot-shuts-down-founder-says-ai-moving-faster-than-regulators/
Orsen Okami
Orsen Okami
https://www.kainjoo.com
Kainjoo is a brand-tech firm serving regulated industries with Kaizen and Six-sigma ready brand activities.

Leave a Reply

Votre adresse e-mail ne sera pas publiée. Les champs obligatoires sont indiqués avec *

Choose country or region

Kainjoo is a group of companies with the sole purpose of bringing brands performances to life in complex industries. We have a global reach with partners and representatives located in all time zones. 

China (Mandarin | English)
Japan (Japanese | English)
Singapore (English)
Australia (English)
India (English)
South Korea (English)

Switzerland (English | French | German)
United Kingdom (English)
France (French | English)
Germany (German | English)
Spain (Spanish| English)
Italy (Italian| English)
Ukraine (Russian| English)

Canada (English | French)
United States of America (English)