Why Woebot Shut Down (And What It Tells Us About AI Therapy)

On June 30, 2025, Woebot Health retired its consumer-facing chatbot. After roughly eight years of operation, what had been one of the most-recommended free AI mental health apps in the world simply turned off. Users got a notice. The app stopped accepting new sign-ups. Existing accounts wound down. The company didn't disappear — it pivoted to B2B partnerships with healthcare organizations — but as far as individual consumers were concerned, Woebot was over.

The shutdown matters beyond the specific app, for two reasons. First: Woebot was the closest thing the AI mental health category had to a category-defining product. Stanford-affiliated, peer-reviewed, widely cited in mainstream media as the example of "responsible AI mental health." When it exited the consumer market, it took a chunk of the category's credibility with it. Second: the reasons it shut down are the reasons most consumer AI mental health apps will face the same decisions in the next few years.

This post is what happened, why, and what it tells us about where the category is going. Full disclosure upfront: ILTY (the publisher) is itself an AI mental health app, which means we're an interested party. Our Woebot alternatives roundup covers what ex-Woebot users can use now. This post is the prior question: what actually happened, and what it means.

The facts

Woebot Health was founded in 2017 by Alison Darcy, a clinical research psychologist who had previously worked at Stanford. The product was a chatbot delivering Cognitive Behavioral Therapy (CBT) techniques through a structured, scripted conversation. The defining design choice was that Woebot was not trying to feel like talking to a human — it was openly a robot, and the robot was clear about that with users.

Three things distinguished it from the rest of the AI wellness category:

1. Research backing. Woebot Health published or co-published more peer-reviewed studies than any other consumer mental health chatbot. The 2017 Fitzpatrick et al. RCT showed reductions in depression symptoms over two weeks. Multiple follow-up studies replicated and extended the work.
2. Scripted, not generative. Woebot did not use the same kind of large language model that drives modern AI chat. The conversation was largely decision-tree-based, with predefined responses authored by clinicians. This was a deliberate safety choice — it meant the chatbot couldn't say something unhinged in response to a vulnerable user, but it also meant the conversations could feel constrained.
3. Free consumer tier. For most of its existence, Woebot had a generous free tier. This drove word-of-mouth, made it a default recommendation in clinician circles, and built a user base in the millions.

On June 30, 2025, that consumer app retired. Woebot Health, the company, continued — but as a B2B service, partnering with healthcare organizations and employers to deliver the chatbot inside their existing care models rather than directly to consumers.

Why: the business model

The shutdown reflects something specific about consumer mental health software that the broader public hasn't fully reckoned with.

Consumer mental health is hard to monetize. The market splits roughly into three tiers:

• Free users — the largest group. Drives word-of-mouth and brand. Often does not convert to paid.
• Subscribers — the middle group, paying $5-15/month. Drives revenue but has high churn; people quit when they feel better.
• High-acuity users — the smallest group, often the ones who need care most. Largely cannot afford recurring subscriptions and often need human care anyway, which the app cannot provide.

The business model problem: the app's primary social good — helping the largest number of people — is concentrated in the free tier, which by definition doesn't generate revenue. The subscribers have the wrong selection (people who can pay and like apps, who often don't have the most severe needs). And the high-acuity users — the people the app was, in some sense, "for" — are economically unviable to serve.

Woebot's specific version of this problem: the most generous free tier in the category, the most clinically credible product, and a paid tier that never grew fast enough to support the company's costs. The product was working clinically. The product was reaching people. The economics didn't close.

The regulatory environment got harder. FDA scrutiny of AI mental health products has intensified through 2024-2025. The question of whether a CBT chatbot is a wellness product or a medical device has gotten sharper, and being on the medical-device side comes with significant cost and slower iteration. For a consumer-facing app trying to ship product fast, this is a hard regulatory posture to sustain. For a B2B partner inside a healthcare system's existing regulatory perimeter, it's manageable.

The competitive landscape shifted. When Woebot launched in 2017, modern conversational AI didn't exist. By 2025, ChatGPT and its peers had created a baseline expectation of free-form natural conversation that scripted decision-tree chatbots couldn't match. Woebot's clinical conservatism — the same conservatism that drove its safety record — read to many users as stiffness compared to LLM-based competitors.

The pivot to B2B isn't a failure of the technology or the clinical work. It's a recognition that the unit economics work better when the buyer is an employer or an insurer paying for an outcome, not a consumer paying for an experience.

What it doesn't say

It's important to be precise about what the shutdown doesn't mean, because the takes after a high-profile exit tend to overshoot.

The technology didn't fail. Multiple peer-reviewed studies show Woebot reduced depression and anxiety symptoms over the periods studied. The clinical methodology is intact. The CBT-via-chatbot approach is still a legitimate tool — it's just no longer available directly to consumers under the Woebot brand.

AI therapy didn't fail. Wysa, Youper, and others continue. The category is growing. The specific business model — free consumer chatbot supported by a small paid tier — is the part that failed, not the underlying premise that conversational software can deliver some elements of CBT and related modalities effectively.

Users didn't fail to adopt it. Woebot reached millions of users. The problem was never engagement. The problem was that engagement and revenue grew on different curves, and the gap closed in the wrong direction.

(For the broader question of whether AI mental health apps work in the first place, see What AI mental health apps get wrong, Can AI help with anxiety? What the research actually shows, and our pillar guide on AI mental health.)

What it tells us about the category

Three implications worth naming.

The "free for everyone" business model in consumer mental health is on borrowed time. Free tiers will still exist, but they'll be loss-leaders to drive paid subscriptions or to feed B2B funnels, not the primary product. Apps that committed hard to free will either pivot or shut down. Apps that built around subscription from day one have a longer runway. Apps with B2B revenue alongside consumer revenue have the strongest position.

Clinical credibility is necessary but not sufficient. Woebot had more research than anyone else in the consumer category. It still couldn't make the consumer model work. The lesson isn't to abandon clinical rigor — it's that clinical rigor doesn't substitute for business-model fit. Apps in the category will need both.

The B2B path is the survival path for many products. Several AI mental health products that started as consumer apps have already followed Woebot's general direction — selling into employer benefits programs, insurance plans, university health systems, or healthcare networks. The user experience changes when the buyer is an organization rather than the user themselves (different priorities, different success metrics), but the company survives. Consumer-only AI mental health is a harder bet to make in 2026 than it was in 2017.

What ex-Woebot users should know

If you used Woebot and are reading this because you're trying to figure out what to do, three concrete pieces of information.

Your data. Woebot Health published a wind-down plan that included data export options for active users. If you didn't take advantage of that at the time and want to follow up, contact Woebot Health directly through their B2B-facing site — they may still be able to help, depending on what they retained.

Your CBT skills. Whatever Woebot taught you about cognitive distortions, behavioral activation, mood check-ins, and the rest is still yours. The skills don't expire with the app. You can keep using them with a paper journal, a different app, or a therapist.

Your next app. The closest replacements depend on what you used Woebot for:

• Free clinically-researched CBT chatbot: Wysa is the closest. Real free tier, multiple published studies, iOS + Android.
• Mood tracking with light CBT: Youper is the strongest. Better mood analytics than Woebot had, paid-only.
• Free peer support (no chatbot): 7 Cups for trained listener conversations.
• Conversational AI (not scripted CBT): Pi, Replika, ILTY — all generative, all different in tone. Our take on which fits what use case is in the alternatives roundup.
• Clinically supervised: MindShift (free, Anxiety Canada — strong nonprofit pedigree) or talk to your doctor about a clinic-supervised CBT program (often available through health plans).

For the full eight-app evaluation with use-case rankings and disclosure, see Best Woebot Alternatives 2026 (After the Shutdown).

Where ILTY stands on this

Honest disclosure: ILTY is building in the same category Woebot was in. We're a consumer mental health app charging a subscription. We watched the Woebot shutdown carefully, and a few things we took from it:

• The economics matter. We started subscription-only (no permanent free tier). That makes our growth slower than Woebot's was, but it means the path to sustainability is real.
• Clinical credibility is the floor, not the ceiling. We're working toward published research the way Woebot did, but we're not assuming research alone makes the business work.
• Differentiation matters more than it used to. When the category was small, generic "AI mental health" was enough positioning. Now there's enough competition that being specifically something — in our case, an alternative to toxic-positivity wellness — matters. Mr. Relentless, the tough-love companion, exists because the gentle-and-validating part of the market is saturated.

We don't claim to have solved the problem Woebot couldn't solve. We're trying to be honest about what the category requires and what trade-offs we've made.

If any of this is useful and you want to try ILTY, the download is at ilty.co/download. If you've decided you don't want a paid app and want a free option, we say so in the roundup — Wysa's free tier is the closest direct replacement for what Woebot offered.

The bottom line

Woebot shut down because the consumer business model didn't close, not because the technology or the clinical work failed. The shutdown matters for the category because Woebot was the most legitimate example of consumer AI mental health that the field had — and its exit signals that the path forward is going to be different than the path that produced Woebot.

For users, the takeaway is simpler. The skills are still real. The category is still functional. The specific app is gone. There are replacements, ranked honestly in our roundup, and the migration is manageable.

Sources & further reading

• Fitzpatrick KK, Darcy A, Vierhile M (2017). "Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial." JMIR Mental Health — the foundational Woebot RCT
• Prochaska JJ et al. (2021). "A Therapeutic Relational Agent for Reducing Problematic Substance Use (Woebot): Development and Usability Study." J Med Internet Res
• APA — Mobile Mental Health Apps Position
• FDA — Software as a Medical Device (SaMD) guidance

Related Reading

• Best Woebot Alternatives 2026 (After the Shutdown): Eight tested replacements ranked by use case.
• Woebot vs Wysa vs ILTY: Honest Comparison: Three-way head-to-head, preserved with a shutdown context note.
• AI Mental Health Apps: The Complete Guide: The pillar — how to evaluate any AI mental health app, with or without Woebot as a reference point.
• What AI Mental Health Apps Get Wrong: The patterns that hurt users across the category, not just at any one app.
• AI Therapy Apps 2026: What Works and What Doesn't: The state of the category as of the year Woebot left it.

ILTY is a mental health support tool, not a substitute for professional care. If you're in crisis, call or text 988.