AI Therapy Companies and Startups to Know in 2026

The AI therapy company landscape in 2026

Companies building AI mental health support sort into four camps. Consumer chatbot startups sell support apps directly to users: Slingshot AI, Wysa, Youper, Abby, Earkick, and a long tail of smaller players. Clinician-side and enterprise companies sell to health systems, employers, and therapists themselves: Limbic, Lyra Health, Spring Health, and the platform companies adding AI features. Big AI labs run the general chatbots where most emotional-support conversations actually happen: OpenAI, Google, Anthropic, and the persona platform Character.AI. And research groups, most visibly the Dartmouth team behind Therabot, build the tools that generate the strongest evidence.

One legal fact frames all four camps: no company can deliver therapy through AI. Several states have restricted AI from acting as a therapist, and every serious product now describes itself as support, wellness, or a tool for licensed professionals. The interesting question in 2026 is which business models survive that boundary, and the answer is becoming visible in where the money moves.

This page maps the players, the funding, and the rules, then closes with a checklist for evaluating any company's claims. For running coverage of new launches and laws, our AI therapy news page tracks developments as they land.

Consumer chatbot startups: crowded, loud, and fragile

The best-capitalized pure play is Slingshot AI, the New York startup behind Ash. It launched Ash publicly on July 22, 2025, after a beta with 50,000 users, backed by $93 million from investors including a16z, Radical Ventures, Forerunner, Felicis, and Menlo. Its founders are AI engineer Daniel Cahn and Casper co-founder Neil Parikh, and its launch branding called Ash the first AI designed for therapy, even as the app itself disclaims being therapy. That tension between marketing and disclaimer runs through the entire consumer lane.

The older clinical-minded players took a different road. Wysa, known for its penguin chatbot and its structured CBT exercises, built a business on employer and health system deals alongside its consumer app. Youper paired its chatbot with validated mood measures. Both hold FDA Breakthrough Device designations, which speed regulatory review but are routinely confused with approval in marketing copy. No AI therapy chatbot holds FDA clearance or approval as a treatment. Below them sits a long tail of small apps, many with anonymous owners, competing on price and search traffic.

The cautionary tale of the category is Woebot. Founded by Stanford-trained psychologist Alison Darcy, it was the most rigorously studied consumer mental health chatbot, and roughly 1.5 million people used it over the years. Its consumer app still shut down, with June 30, 2025, as its last day. Darcy pointed to the cost and difficulty of pursuing FDA marketing authorization while large language models raced ahead of the regulatory framework, and the company pivoted to enterprise work. Our Woebot review tells the full story.

The lesson investors took from that shutdown is uncomfortable for the whole lane: the app with the best evidence in the category could not make consumer economics work. Free general chatbots cap what users will pay, churn is high, and regulation keeps raising the cost of doing it right.

Clinician tools and enterprise platforms: where the money moved

The quieter camp sells AI that supports licensed humans instead of replacing them. Limbic built clinical triage and intake AI used across NHS talking-therapy services in the UK. Lyra Health and Spring Health embed AI in care navigation and matching for employer mental health benefits. Talkspace, a human-therapist platform, has been building AI features around its clinicians rather than in place of them. In this camp, the AI handles screening, routing, documentation, and between-session support, and a licensed professional stays responsible for care.

Consolidation is the 2026 story here. Spring Health announced on January 29, 2026, that it would acquire Alma, the membership platform that helps independent therapists accept insurance, and the deal closed on May 1, 2026. The combined company supports more than 170 million covered lives across employers and health plans. The direction is clear: the durable businesses are building networks and infrastructure around human clinicians, with AI as the connective tissue.

The structural advantages of this camp explain the shift. Selling to employers and health plans produces contract revenue instead of consumer churn. Using AI for triage and support instead of standalone therapy sidesteps the new state bans entirely. And outcomes can be measured and reported to the payer, which consumer apps rarely manage. When people ask where AI integration in digital therapy is heading, this is the honest answer: into the workflow around the therapist, faster than into the chat window itself.

Big AI labs and research groups

The largest mental health chat surface in the world is not a mental health company. OpenAI disclosed in October 2025 that about 0.15 percent of ChatGPT's 800 million weekly users, roughly 1.2 million people, talk to it about suicide in a given week. Its October 29, 2025, model update, built with input from more than 170 psychiatrists and psychologists, aimed to make the model respond more safely in conversations involving distress, psychosis, and emotional reliance. Those numbers explain why regulators treat general chatbots as mental health infrastructure whether or not the companies market them that way.

Character.AI shows the liability side of that reality. Users built therapist-styled personas on its platform, minors formed intense attachments, and families sued after teen suicides. On January 7, 2026, news broke that Google and Character.AI had agreed to settle multiple lawsuits, including the case brought by Megan Garcia, whose 14-year-old son Sewell Setzer III died by suicide after extensive conversations with the platform's chatbots. The settlements closed the first major legal chapter on AI companion harm, and the litigation map it leaves behind now shapes every product decision in the category.

The research camp moves slower and matters more than its size suggests. Dartmouth's Therabot, developed over years by clinicians and engineers, delivered the field's most credible result: a randomized trial with 210 participants, published March 27, 2025, in NEJM AI, showing meaningful symptom reductions for depression and anxiety versus a waitlist. Groups like this set the evidence bar the commercial players get measured against, and the gap between research-grade development and startup pace remains the most underrated fact in the market.

The regulatory headwinds reshaping the market

State law is now the strongest force acting on these companies. Illinois Governor JB Pritzker signed the Wellness and Oversight for Psychological Resources Act on August 1, 2025, banning AI from delivering therapy on its own, with civil penalties up to $10,000 per violation. Nevada enacted its own restriction on AI providing mental health services, and Utah requires mental health chatbots to disclose that users are talking to software. Disclosure rules are spreading through states including California, Colorado, and Maine, and more mental health bills are moving. A national app has to comply with the strictest state it operates in, which effectively makes Illinois the floor. Our breakdown of the Illinois ban covers what the law actually restricts.

Federal scrutiny arrived in parallel. On September 11, 2025, the Federal Trade Commission issued 6(b) orders to seven companies, including Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap, and X.AI, demanding detail on how their companion chatbots measure and mitigate harm to children and teens. The inquiry remains active in 2026, and its findings will likely shape the first federal rules for this category.

The practical effect on company behavior is visible everywhere: the word therapy is disappearing from product marketing, crisis routing and AI disclosure are becoming standard features, and legal budgets are becoming a startup cost. That pressure favors the well-funded and the clinician-aligned, which is one more reason the enterprise camp is pulling ahead. Where all of this lands is the subject of our look at the future of AI in therapy.

How to evaluate any AI therapy company's claims

Start with people. Look for named, licensed clinicians on the team or advisory board, with credentials you can verify. Remarkably few companies in this market name a single one, including some that market themselves as built by therapists. A company that will not tell you who its clinicians are is answering your question by omission.

Then check the evidence and the safety mechanics. Published, peer-reviewed research on the actual product is the gold standard, and press releases about internal studies are not that. Test the crisis response yourself: a trustworthy tool tells you clearly that you are talking to AI, surfaces crisis resources when the conversation darkens, and encourages human help. If you are in crisis or thinking about suicide, call or text 988 (US Suicide & Crisis Lifeline), available 24/7. No company on this page offers a product built for that moment.

Finally, read the business model and the company's survival odds. Who pays, and what does the company do with your conversation data? A free product with no visible revenue is monetizing something, and it may be you. Funding is a signal to read carefully in both directions: a big raise funds growth ahead of evidence, while a thin balance sheet raises the risk your support tool disappears the way Woebot's app did, taking your routine and your data relationship with it. Companies come and go quickly in this market. Your standards for evidence, transparency, and privacy should not.