On April 24, 2026, Grow Therapy rolled out "Coach" — an AI chat feature that sits with your clients between sessions, drawing on CBT, DBT, ACT, and Behavioral Activation. Since the December 2025 pilot, more than 800,000 messages have been sent, and 50% of active Grow Therapy providers already have at least one client using it (PRNewswire, April 24, 2026). That number moved fast. And whether you're a Grow Therapy provider or not, the implications reach into any solo practice that exists inside a platform-dependent care model.
What Grow Therapy's AI Coach Actually Does
Grow describes Coach as an AI-powered support tool for clients in the 26 days between their four sessions each month. It's not a diagnosis tool. Officially, it's positioned to extend therapeutic momentum between sessions, not replace the clinician. The company says it uses evidence-based frameworks — CBT, DBT, ACT, Behavioral Activation — and launched after a testing period that began in December 2025.
Grow's scale matters as context here. As of March 2026, the company serves 26,000 providers, carries a $3B valuation after a $150M Series D, and reaches 220 million Americans through 125+ health plan partnerships (Bloomberg, March 3, 2026). When a company that size ships a feature, it isn't a pilot you can scroll past. It's a structural decision about how care flows on their platform.
The product is also a bet on a specific regulatory gap. The FDA's January 2026 guidance expanded the "general wellness" exemption, which means an AI chat tool that avoids diagnostic claims likely falls outside FDA medical device oversight (Nixon Law Group, January 2026). Coach was designed to live in that gap. That's not an accident. That's a product strategy.
What "Clinician Oversight" Actually Means
Grow calls this a "clinician oversight" model, and that framing deserves a close read. The structure is more layered than the press release suggests.
By default, therapists can see client chat messages. Both the therapist and client must consent before Coach activates. Therapists can disable it for any individual client at any time. A clinical quality team reviews safety flags and random message samples, and Grow claims a 99% safety concern detection rate (PRNewswire, April 24, 2026).
Here's what matters: clients can activate "private mode." When they do, their messages are hidden from the therapist — except for safety flags. So your visibility into what the AI says to your client, and how your client responds, is opt-out for the client, not locked-on for you.
That's a meaningful architectural choice. It isn't a scandal. But it is a reason to read the consent documentation before the first client opts in, rather than after.
What Went Wrong Before — and Why Grow's Stanford Partnership Matters
This isn't the first time AI has occupied the space between sessions. The past 18 months have shown what happens when that space has no clinical floor.
Woebot, which raised $123 million, shut down its consumer app on June 30, 2025. The company cited FDA regulatory costs and the absence of a clear LLM rule framework. That's not a failure of AI in mental health generally. It's what happens when a product operates without a clinical anchor and the regulatory ground hasn't been defined yet (STAT News, July 2, 2025).
Character.AI settled multiple teen suicide lawsuits in January 2026. Those weren't mental health products — they were companionship chatbots — but the legal fallout hit the entire AI-in-mental-health category. APA's 2026 data reflects where clinicians have landed: 89% of psychologists worry AI chatbots may encourage self-harm, 85% worry they'll pose as licensed therapists, and 94% do not trust tech companies to protect patients' mental health data (APA, 2026 Chatbots Mental Health Report).
Grow appears to know the credibility gap is real. On June 23, 2026, they announced a research partnership with Stanford University to establish clinical safety standards for AI in mental health, led by PI Jonathan Chen, MD. The inaugural study measures how leading AI models respond to mental health crises (PRNewswire, June 23, 2026).
That partnership was announced after the product launched, not before. The research is catching up to the deployment. That sequencing — ship first, build the safety evidence second — isn't unique to Grow. It's the pattern across almost every AI mental health product. Clinicians who wait for the research before forming a policy will always be reacting to something already in use.
What This Means for Your Solo Practice
The platform-dependency question here is distinct from the AI-safety question, and it's the one most solo clinicians aren't asking yet.
If you're a Grow Therapy provider, a third party is now available to communicate with your clients between sessions. You can disable it per client. But it defaults to active once a client opts in, and the consent architecture is Grow's, not yours. The practical question isn't whether Coach is good or bad. It's whether you've reviewed the consent flow deliberately, or whether you'll find out what a client shared with the AI when it surfaces in a session.
If you're not on Grow Therapy, this is still a signal. Every managed-care-adjacent platform is watching this launch. The AI between-session product is going to become a standard feature across the industry within the next two years. Your relationship with any network or platform will eventually include some version of this.
What's worth distinguishing: AI that helps you work, versus AI that operates independently in your clinical relationship. Notes summarization, treatment planning support, session prep — those tools serve the clinician. A chat feature your client uses when you're not available serves the platform's retention and engagement goals, among other things. Both can be clinician-friendly. They're not the same category of tool.
There are clinician-built tools designed around the first model. VibeCheck.luxury is one example: built by a licensed clinician, designed to support the therapist's work rather than occupy the between-session space independently. It's worth knowing the distinction exists when you're evaluating what belongs in your practice.
What to Actually Do Right Now
This isn't a call to leave Grow Therapy or to reject AI in mental health. It's a call to make a deliberate decision before one gets made for you.
If you're a Grow Therapy provider, read the Coach consent documentation now. Understand the private mode option and what it hides from you. Decide per client whether you want Coach active. That control exists — use it before your clients discover the feature on their own.
If you're evaluating platforms, add this question to your checklist: how does this platform handle AI between-session contact, and what are my consent rights as the treating clinician? The answer will tell you a lot about how the platform thinks about clinical authority.
The question on AI tools generally isn't whether they belong in mental health. The question is who controls them, who they report to, and what happens when something goes wrong.
FAQ
What is Grow Therapy's AI Coach and when did it launch?
Grow Therapy launched Coach on April 24, 2026, after a testing period beginning December 2025. It's an AI chat feature for clients between sessions, using CBT, DBT, ACT, and Behavioral Activation. As of the launch announcement, more than 800,000 messages had been sent, and 50% of active Grow Therapy providers had at least one client using it (PRNewswire, April 24, 2026).
Can therapists see what clients say to Grow Therapy's AI Coach?
By default, yes. Grow's model shows therapists their client's Coach messages. However, clients can activate "private mode," which hides messages from the therapist except for safety flags. Therapists can also disable Coach for any individual client at any time, and both parties must consent before it activates.
Do therapists have to participate in Grow Therapy's AI Coach?
No. Consent from both the therapist and the client is required before Coach activates for any given client. Therapists can disable it for specific clients at any time. The control structure exists — the practical task is knowing how it works before clients start opting in on their own.
What happened with previous AI mental health tools like Woebot?
Woebot shut its consumer app on June 30, 2025, after raising $123 million, citing FDA regulatory costs and unclear LLM guidance. Character.AI settled multiple teen suicide lawsuits in January 2026. APA's 2026 Chatbots Mental Health Report found 89% of psychologists worry AI chatbots may encourage self-harm, and 94% do not trust tech companies to protect patients' mental health data (APA, 2026).
Is Grow Therapy's AI Coach regulated by the FDA?
As of mid-2026, Coach likely falls outside FDA medical device oversight. The FDA's January 2026 guidance expanded the "general wellness" exemption to cover AI tools that avoid diagnostic claims (Nixon Law Group, January 2026). Grow Therapy announced a Stanford University research partnership in June 2026 to develop clinical safety standards, but that research is ongoing.