The AI Therapy Gold Rush Has a Regulatory Problem

The AI therapy market is projected to hit $114 billion by 2035, growing at 27.8% CAGR. Over 40 million people now use AI-powered mental health apps monthly. Investors have poured billions into the space.

And in the middle of this gold rush, one of its most well-funded companies just collapsed.

The FDA’s January 2026 updated guidance on general wellness products broadened exclusions for low-risk digital health tools, creating a clearer regulatory path for AI-powered wellness applications that don’t make medical claims.

The Woebot Story

Woebot Health raised over $100 million. It had 14 randomized controlled trials — more clinical evidence than almost any competitor. It received FDA clearance for treating adolescents with depression.

Then in early 2026, it shut down.

The founder’s explanation was blunt: regulatory limbo. The FDA had a clear pathway for rule-based AI systems, but no established framework for large language models. Woebot couldn’t get the guidance it needed to operate within the law, so it stopped operating entirely.

This wasn’t a failure of product or market. It was a failure of regulatory infrastructure to keep pace with technological capability.

The Regulatory Patchwork

The Woebot story is one data point in a broader pattern. Across the US, the regulatory landscape for AI mental health tools is fragmented and inconsistent:

Federal level: The FDA updated its general wellness guidance in January 2026, broadening exclusions for low-risk wellness apps. But the line between “wellness” and “medical” remains unclear for AI systems that can detect emotional tone, adapt therapeutic techniques, and maintain long-term memory. The FDA has established a Predetermined Change Control Plan pathway for AI/ML devices, but its application to LLM-based systems is still being developed.
State level: Illinois passed a law banning AI from making therapy decisions, with $10,000 fines per violation. Other states are considering similar legislation. The result is a compliance nightmare for any AI therapy company operating nationally.
FDA Breakthrough pathway: Wysa, a text-based AI therapy platform, secured FDA Breakthrough Device designation for its conversational agent. This provides a faster path to market, but it’s specific to Wysa’s bounded clinical scope — it doesn’t create a general pathway for the category.

The gap between what’s technologically possible and what’s legally clear is where companies are getting stuck.

What Actually Needs Regulation

The regulatory confusion stems from a category problem: not all “AI therapy” is the same thing, but the regulations treat it as a monolith.

Consider three distinct categories:

Category 1: Diagnostic and treatment AI Systems that make clinical decisions — diagnosing depression, recommending treatment protocols, adjusting medication. These clearly require FDA oversight. Woebot’s CBT chatbot arguably fell into this category for some use cases.

Category 2: Guided self-help AI Systems that deliver structured therapeutic exercises within a defined scope. Wysa’s CBT exercises and coaching fall here. This is where FDA Breakthrough designations have been granted.

Category 3: Wellness and performance AI Systems that support general well-being, stress reduction, and performance optimization — without making clinical claims or delivering medical treatment. The FDA’s 2026 guidance update broadened exclusions for this category.

Oriamind operates in Category 3. The sessions are not medical treatment. They are structured guided hypnosis and visualization sessions for performance, mindset, recovery, and sleep — the same category as guided self-hypnosis apps that have existed for decades, now enhanced by AI personalization.

The distinction matters because regulatory clarity is essential for the industry to function. Companies need to know which category they’re in and what rules apply.

What the Woebot Shutdown Means

The Woebot shutdown is not a sign that AI therapy is failing. It’s a sign that the regulatory framework needs to catch up.

Key implications:

Companies with strong clinical evidence are not immune — Woebot had 14 RCTs and FDA clearance. It still couldn’t navigate the regulatory uncertainty around LLMs.
State-level fragmentation is a business risk — if every state writes its own AI therapy rules, national deployment becomes impractical.
The “wellness” category is the safe harbor — companies that stay within wellness and performance claims have clearer regulatory footing. The FDA’s 2026 guidance explicitly broadened exclusions for low-risk wellness products.
Integration models will win — the companies that succeed will be those that combine AI efficiency with human oversight, operating within established regulatory frameworks rather than trying to create new ones.

The Market Reality

Despite the regulatory uncertainty, the digital therapeutics market continues to grow rapidly:

$12.58 billion in 2026
Projected $114.35 billion by 2035
27.8% CAGR
40 million+ monthly active users of AI therapy apps
$8 billion mental health tech market

Spring Health reached a $3.3 billion valuation. Wysa continues to operate with FDA Breakthrough designation. New companies enter the space weekly.

The growth is real. The user demand is real. The clinical evidence — for appropriate use cases — is real.

What’s missing is regulatory clarity that allows companies to build durable, compliant businesses within the category they actually operate in.

The Bottom Line

The AI therapy gold rush isn’t over. It’s entering a phase where regulatory strategy matters as much as product strategy. Companies that understand their category, operate within established frameworks, and build for long-term compliance will outlast those that try to move fast and break things.

Mental health is not the domain for “move fast and break things.” It never was. The winners will be the ones who understood that from day one.

Adam Shaaban is the founder of Oriamind. LinkedIn · X / Twitter

How to Apply This

If you’re building or investing in AI therapy:

Know which category you’re in — diagnostic/treatment AI, guided self-help, or wellness/performance. Each has a different regulatory path.
Stay within your claims — the companies that survive are those whose product category matches their regulatory strategy.
Build for the FDA’s 2026 guidance — the broadened exclusions for low-risk wellness products create a clear path for companies that don’t make medical claims.
Plan for state-level fragmentation — Illinois banned AI from therapy decisions. More states will follow. National deployment requires navigating a patchwork.

This article is part of our AI hypnotherapy & behavioral change series.