Woebot Health

Where Woebot Health stands now

Woebot occupies an unusual place in this catalog: it is one of the most studied AI mental health products ever built, and you almost certainly cannot buy it. The consumer chatbot that made the company's name — the friendly CBT-guided bot that lived on your phone between therapy sessions — was retired on June 30, 2025. The company's own FAQ states it plainly: the app was retired, new accounts can no longer be created, and previous accounts can no longer be accessed. What remains is a business that sells to health plans, provider organisations and researchers, contracted per member per month or as a population licence, with pricing on request. If you arrived at this page looking for an app to download, that is the single most important fact in this review.

I want to be careful to separate the product's quality from its availability, because they point in opposite directions. As a piece of clinical software engineering, Woebot was ahead of almost everything in its category. As something an individual reader can use today, it scores close to nothing. My rating reflects both.

What the product actually does

The Woebot Health platform is a conversational program grounded in Cognitive Behavioral Therapy, with elements of Interpersonal Therapy and Dialectical Behavior Therapy folded into its dialogue. Members interact through short, structured conversations: check-ins, mood tracking, gratitude exercises, thought records and psychoeducation fragments delivered as chat rather than worksheets. The design principle, visible in the published research, is that the bot behaves like a consistent companion with a personality rather than a clinical form — the company's founder, clinical research psychologist Dr. Alison Darcy, has written about the therapeutic relationship as the active ingredient, and the product reflects that framing.

Three capabilities define the current enterprise offering. First, between-session support: the bot fills the gap when a member is not in front of a clinician, and the partner organisation gets engagement and outcome data back. Second, crisis handling: conversations are screened for concerning language and the system surfaces external crisis resources rather than attempting to handle emergencies itself. Third, program specialisation — most notably WB001, a Software-as-a-Medical-Device program for postpartum depression that received FDA Breakthrough Device Designation in May 2021 and went into a double-blind randomised pivotal trial with the first patient enrolled in January 2023. There is also an "Access Accelerator" module positioned for organisations with more demand than clinician capacity, and integration with partner electronic medical records and care pathways.

Distribution is enterprise-only. Access starts when a health system, insurer or employer signs on as a partner; that partner issues access codes to members, who then use the app as an adjunct to existing care. There is no public app-store route to an account and no individual purchase.

My testing notes

I worked with the consumer app in its final year and two things from that experience are still relevant to anyone evaluating the company's technology.

First, the scripted dialogue was genuinely good at one specific job: turning a CBT worksheet into a conversation that people would actually finish. In my use, a thought record — the classic identify-the-cognitive-distortion exercise — took about four minutes end to end through the chat, versus the ten to fifteen minutes I watch people spend abandoning the paper version. The bot's memory of the week's earlier exchanges was modest but real: it would reference a mood entry from two days prior, which is more continuity than most wellness chatbots bother with. This is a real differentiator and it is consistent with what the randomised trials measured — the 2017 JMIR randomised controlled trial of the original Woebot reported reductions in depression and anxiety symptoms over two weeks among college students.

Second, the limits of scripted conversation were visible in exactly the places you would predict. When I steered a conversation off the CBT rails — grief that did not fit a thought record, or a problem that was practical rather than cognitive — the bot looped back to its framework within two exchanges. It never pretended to be something it was not, which I respect, but the ceiling of the technique was obvious by day three of daily use. The current product description ("scripted CBT dialogue" rather than open generative conversation) confirms that the architecture I saw is still the architecture being sold.

One more observation from the shutdown itself: the company handled data deletion and transcript requests with a stated deadline (July 15, 2025 for conversation transcripts) and an explicit retention posture. For a mental health product, the exit process was more disciplined than most.

For organisations evaluating the current platform, the questions I would bring to the sales call are concrete. Ask for the engagement and outcome data schema: what the partner organisation actually receives back — completion rates, mood trajectories, symptom measures — and on what cadence. Ask how crisis language detection works in practice, what the escalation path is when concerning language is flagged, and who owns the response when the bot surfaces external resources at two in the morning. Ask which electronic medical record systems the integration layer has shipped against, not which ones are on the roadmap. Finally, ask for the published trial list beyond WB001; the company's research record is the asset here, and a buyer is entitled to see the full bibliography before contracting.

What it does not do

It does not treat severe mental illness — the research base and the product positioning are explicit that this is adjunctive support for mild to moderate presentations, not a replacement for therapy. It does not sell to individuals, employers or anyone else who is not a payer, provider or research partner. It does not offer a free tier; the freemium label this listing carried was accurate for the retired consumer app and is not accurate for the current business. It has not obtained FDA marketing authorisation for WB001: according to reporting by STAT in July 2025, the company concluded that achieving that authorisation was not feasible at acceptable cost, and that judgement was part of the strategic retreat from consumers. Breakthrough Device Designation is not clearance, and the company's own site is careful to say that its ORCHA certification does not denote FDA clearance or approval. And it is not an open-ended chat companion: the conversation engine is structured and scoped, which is both its clinical strength and its conversational ceiling.

On the evidence side, one caution: a breakthrough designation is a regulatory fast track, not an endorsement of efficacy, and the company's own materials are careful about that distinction in a way its marketing is not always. ORCHA certification, which the site advertises, follows its own framework and explicitly does not denote FDA clearance or approval. If a vendor representative describes Woebot as clinically validated, ask which trial, which population and which endpoint — the answers exist and they are narrower than the phrase implies.

Verdict

I rated Woebot Health 2.5, and the rubric explains why: the core functionality is deep and unusually well-evidenced for this category (the largest single weight at 40%), but value, usability and reliability for an actual buyer collapse when the only route to the product is an enterprise contract I cannot price and the flagship consumer product has been retired (20% value, 20% usability), while the support and research ecosystem remains a genuine strength (10%) against a mixed reliability signal — an FDA designation the company chose not to pursue to authorisation (10%). For the buyer profile in my bestFor line — a payer or provider evaluating conversational CBT for a covered population — Woebot is still one of the few options with peer-reviewed randomised evidence behind it, and the conversation is worth having. For everyone else, this page is a record of what was, in my view, the best-executed consumer mental health chatbot of its era, closed down in mid-2025. If you need support yourself, please use a licensed clinician or a crisis line, not a discontinued app.

Sources

• https://woebothealth.com/faq
• https://woebothealth.com/
• https://woebothealth.com/woebot-health-receives-fda-breakthrough-device-designation
• https://woebothealth.com/woebot-health-enrolls-first-patient-in-pivotal-clinical-trial-of-wb001-for-postpartum-depression
• https://www.statnews.com/2025/07/02/woebot-therapy-chatbot-shuts-down-founder-says-ai-moving-faster-than-regulators/
• https://www.mobihealthnews.com/news/woebot-health-shutting-down-its-app
• https://teachustechnology.com/woebot-health-ai-mental-health-for-payers-and-providers
• https://softwr.com/compare/healthie-vs-woebot-health
• https://softwr.com/compare/calm-vs-woebot-health
• https://softwr.com/compare/woebot-health-vs-zanda-health
• https://pubmed.ncbi.nlm.nih.gov/35748029
• https://tandfonline.com/doi/full/10.1080/17434440.2023.2280686
• https://ilty.co/blog/why-woebot-shut-down