AI therapy: which startup is ahead?
In our mental health market deck, you will find everything you need to understand the market
SUMMARY
Wysa is ahead overall today. Limbic is almost level in clinical care, while Slingshot AI’s Ash is the strongest consumer challenger.
Wysa’s lead does not come from one spectacular metric. It comes from combining more than six million users, institutional distribution, a large research library, coaching, referrals and newer routes into primary care and physical therapy.
Limbic is already ahead where AI therapy becomes formal healthcare infrastructure. Its NHS penetration, medical-device certification and integration into assessment and treatment workflows are harder to reproduce than a polished chatbot experience.
The user numbers need careful reading. Wysa and Youper report cumulative reach, The Path reports active members, and Limbic counts clinical assessments, so the largest headline number is not automatically the strongest operating metric.
Ash appears to be adding users fastest, moving from a beta population of about 50,000 to roughly 200,000 people. The catch is retention: more than half of the accounts in its own study dropped out early, and only a small fraction became heavy users.
Limbic has produced the clearest measurable customer value. Its published work points to shorter assessments, faster access to care, lower dropout and better treatment outcomes, which gives healthcare buyers a much stronger business case than download or conversation totals.
Limbic now has the strongest recent clinical evidence, while Wysa has the broadest research base accumulated over time. Ash and The Path have interesting early datasets, but their evidence still says more about engagement and safety than proven treatment outcomes.
Regulation is likely to widen the gap between clinical and consumer-first companies. Limbic and Jimini were designed around professional oversight, while products that sell an autonomous therapeutic relationship may eventually have to narrow their claims or build much heavier compliance systems.
Capital efficiency sharply favors Wysa and Limbic. Ash has raised $93 million, yet Wysa has reached far more people with roughly one-third of that funding, and Limbic has embedded itself across much of NHS Talking Therapies with about $14.7 million.
AI therapy is splitting into three markets: entry into care, support between professional sessions, and direct consumer relationships with an AI. Wysa leads overall because it touches all three, but a successful US expansion could put Limbic first, and stronger retention or paid demand could quickly lift Ash.
This market map, featured in our mental health market deck, highlights top companies and startups in the mental health market
AI therapy: which startup is ahead?
Which AI therapy startups count today?
Eight independent startups deserve serious comparison today: Wysa, Limbic, Slingshot AI’s Ash, Jimini Health, The Path, Youper, Earkick and Sonia.
We include companies whose AI speaks directly with people about their mental health, either independently or as part of professional care. That gives us three different groups: consumer AI companions such as Ash and The Path, established self-help apps such as Wysa and Youper, and clinical systems such as Limbic and Jimini.
We leave out ChatGPT, Claude and Gemini because they are general-purpose assistants. Replika and Character.AI are primarily companion products. BetterHelp, Talkspace and similar platforms still deliver therapy through human professionals, even when AI helps with matching, notes or administration.
Woebot would have made this list a year ago. Its consumer app closed in 2025 after reaching around 1.5 million people and raising close to $100 million. Clare&me also disappeared as an independent competitor after an insolvency process and asset sale. Their exits show how difficult it is to build a lasting business, even with users, funding and published research.
Yuna and several smaller wellness bots sit just outside our comparison. They offer emotional support, but the public evidence is still too thin to judge them against companies already operating at meaningful consumer or clinical scale.
Funding figures for private startups remain approximate. We use announced rounds and the most consistent totals across company statements and established funding databases.
| Startup | What it currently does | Reported funding |
|---|---|---|
| Slingshot AI, Ash | Generative AI built for long, personalized mental-health conversations | $93 million |
| Wysa | AI self-help combined with coaching, employer benefits and clinical pathways | About $30 million |
| Jimini Health | Clinician-supervised AI support before, after and between therapy sessions | More than $25 million |
| Limbic | AI assessment, referral and CBT support inside healthcare services | About $14.7 million |
| The Path | Consumer AI therapy and coaching with persistent memory | $14.3 million |
| Youper | Established AI wellbeing app using structured psychological exercises | About $5.2 million |
| Earkick | AI self-care companion with mood and anxiety tracking | About $1 million |
| Sonia | Voice and text companion for emotional reflection | About $500,000 |
Is one AI therapy startup clearly ahead right now?
Wysa leads AI therapy overall today, although Limbic is close enough to win whenever we focus strictly on regulated clinical care.
Wysa has the most complete business. It reaches consumers directly, works with employers and insurers, supports NHS pathways and now extends into primary care and physical therapy. Its current disclosures include more than six million users, one billion AI conversations, 11 million covered lives and over 45 peer-reviewed publications.
Limbic has gone deeper into healthcare. Its products have delivered more than 650,000 clinical assessments and have been used by about 45% of NHS Talking Therapies services. The company also holds UK Class IIa medical-device certification for Limbic Access.
Wysa leads on reach, distribution and product breadth. Limbic leads on clinical penetration, recent research and integration into treatment pathways.
Ash sits in third place. It has the largest funding pool, a purpose-built behavioral-health model and roughly 200,000 reported users, but it still discloses little about paid demand, retention or healthcare adoption.
The Path and Jimini are the two newer challengers worth watching. The Path reports more than 50,000 active members and 2.5 million sessions, while Jimini has raised over $25 million to deploy clinician-supervised AI.
If you want more recent data on this point, please see our latest mental health market report.
As this chart shows, and as featured in our mental health market deck, search interest in men’s mental health has been rising steadily
Who has the most AI therapy users today?
Wysa still has the largest disclosed AI therapy audience, with roughly twice the cumulative reach reported by Youper and about 30 times Ash’s current user count.
Wysa reports more than six million users. Youper’s long-standing figure is approximately three million, while Ash says it has reached around 200,000 people. The Path currently reports more than 50,000 active members.
These figures measure different things. Wysa and Youper publish cumulative user totals. The Path uses “active members,” although it does not explain the activity window. Limbic’s 650,000 figure measures assessments completed inside healthcare pathways.
Wysa’s billion reported conversations equal around 167 conversations per user. Usage will be concentrated among a smaller group, but the total confirms that its scale goes beyond downloads.
The Path reports 2.5 million sessions across more than 50,000 active members, or about 50 sessions per member. Its definition of a session is broader than a clinical appointment, but the figure still suggests repeated use.
Youper remains second by cumulative reach, although its lack of fresh activity data makes the current strength of that audience difficult to judge.
Who is growing fastest in AI therapy now?
Ash is currently adding users fastest. Limbic is producing the more valuable kind of growth inside real healthcare services.
Ash launched publicly after a beta involving around 50,000 people. Its chief executive later said the app had reached roughly 200,000 users, representing a fourfold increase from the beta population.
Retention is the missing part. An Ash research preprint examined 102,684 accounts and found that 52.2% belonged to an early-dropout group. Only 1.6% qualified as power users, while 25.3% returned weekly.
Limbic said it had helped more than 260,000 NHS patients enter care when it announced its Series A. It now reports over 650,000 assessments. The definitions differ, but the company is clearly handling several hundred thousand more clinical interactions than it was two years ago.
Wysa has moved from about 4.5 million users in 2022 to more than six million currently. Its percentage growth is slower, although it started from a much larger base.
The Path already reports 50,000 active members and 2.5 million sessions, but we cannot calculate a clean recent growth rate because the company inherited earlier usage from its previous identity, Mental.
Ash wins on raw user growth. Limbic’s expansion carries more commercial weight because it comes through healthcare deployments.
This chart, featured in our mental health market deck, illustrates yearly VC investment in mental health startups
Who has turned AI therapy into real customer value?
Limbic has created the clearest measurable value for healthcare customers. Wysa has built the broadest commercial business across providers, employers, insurers and consumers.
Private companies in this market rarely disclose revenue or contract values, so we have to judge demand through deployments, repeat use and the amount of work customers entrust to the product.
Limbic has the strongest clinical proof. About 45% of NHS Talking Therapies services have used its technology, and Limbic says it integrates with 95% of the patient-management systems used in that market. Its product supports referral, assessment, treatment and documentation.
Its economic evidence is also the strongest. A study covering 64,862 patients found that assessments took 12.7 minutes less, patients were seen five days sooner and dropout fell by 18%.
A separate study of 244 group-CBT patients found a 23-percentage-point lower dropout rate, a 25-point higher recovery rate and a 21-point higher reliable-improvement rate among users of Limbic’s between-session tool. Patients chose whether to use it, so the groups were not perfectly comparable.
Researchers estimated that the 21-point increase in reliable recovery could generate about £228 of value per patient using standard NHS treatment-cost assumptions. It remains an early estimate, but competitors have published little comparable work.
Wysa has a wider set of buyers. It works with NHS services, payors, employers and public-health programs, including Singapore’s mindline.sg platform. Its referral technology reports a 91% completion rate and around 30 minutes saved per assessment.
Across 100,000 referrals, half an hour per assessment would release approximately 50,000 staff hours. The exact saving depends on how each service reorganizes the work, but the scale is easy for buyers to understand.
Wysa has also acquired April Health and Kins, giving it routes into American primary care and physical therapy. Wysa increasingly looks like a care platform rather than a standalone chatbot.
Jimini is pursuing a similar clinical model in the United States. Its Sage assistant works under a clinician’s supervision, but the company has not named enough customers or disclosed deployment volumes.
Consumer pricing tells us less about value. The Path charges $40 per month for its Pro plan, while Ash launched free and has not published a clear long-term pricing model. Both still need to show strong paid retention.
If you want more recent data on this point, please see our latest mental health market report.
Who has the strongest clinical evidence?
Limbic now has the strongest clinical evidence in AI therapy. Wysa has built the largest and most varied research library over time.
Limbic’s strongest result comes from a 2026 Nature Medicine study. In a randomized, double-blind evaluation, 227 people held natural conversations with different therapy agents, and 22 clinicians rated the transcripts. Limbic’s cognitive-layer system scored above standalone language models and the participating human clinicians across the study’s measures of CBT competence.
The researchers then examined 19,674 real-world transcripts from 8,920 users. Greater activation of Limbic’s specialized reasoning system was associated with stronger symptom improvement and a higher chance of recovery around ten weeks later. Several authors are connected to the company, and the real-world analysis cannot prove causation on its own.
Earlier research gives Limbic unusually large clinical datasets. A Nature Medicine study covering 129,400 NHS patients found that services using Limbic’s referral chatbot increased referrals by 15%, compared with 6% in matched control services. The rise was particularly large among groups that had previously been underrepresented in treatment.
Another peer-reviewed analysis followed 64,862 patients across nine NHS services. It found shorter assessments, five fewer waiting days, 18% lower dropout and 45% fewer changes to the assigned treatment pathway.
Wysa’s evidence covers more products and populations. The company lists over 45 peer-reviewed publications and received an FDA Breakthrough Device designation for a narrowly defined chronic-pain program involving depression and anxiety. That designation does not mean the general Wysa chatbot has full FDA approval as a therapist.
Wysa also recently published a study involving 1,213 coaching users. People who had AI support between coaching sessions completed nearly three times as much of their program as people receiving coaching alone.
Ash has gathered the largest recent consumer dataset. Its study of 102,684 users found reductions in self-reported depression and anxiety among the much smaller group who completed follow-up measures. Only 298 people completed both the initial and three-week PHQ-9 and GAD-7 assessments.
The Path’s strongest public result concerns safety. It says its model scored 95 on VERA-MH, compared with a highest score of around 65 among consumer chatbots. The benchmark focuses on responses to suicidal ideation rather than treatment outcomes.
Youper’s best-known study followed 4,517 users and found lower anxiety and depression scores after two weeks. It had no randomized control group. Sonia, Earkick and Jimini have yet to publish comparable product-level results.
| Startup | Strongest useful evidence | What it supports | Main limitation |
|---|---|---|---|
| Limbic | Randomized expert evaluation plus large NHS studies | Clinical reasoning, access, efficiency and treatment support | Several studies include company-linked researchers |
| Wysa | More than 45 publications and several controlled studies | Broad evidence across self-help, coaching and specific care pathways | Evidence covers several product versions and use cases |
| Ash | Naturalistic study of 102,684 accounts | Real-world engagement patterns and possible symptom improvement | Follow-up symptom sample was only 298 people |
| The Path | Company-reported VERA-MH score of 95 | Strong crisis-response safety performance | Measures safety in suicidal-ideation scenarios, not efficacy |
| Youper | Observational study of 4,517 users | Possible short-term anxiety and depression improvement | No randomized control group |
| Jimini, Earkick and Sonia | Early studies or limited public validation | Product plausibility | Too little independent evidence for a firm judgment |
This chart, featured in our mental health market deck, shows why Talkspace is winning in mental health
Which AI therapy product is mature enough for real care?
Wysa is the most mature overall product, and Limbic is currently the most mature system inside formal mental-health services.
Wysa has built a global consumer product, enterprise delivery, human coaching, clinical referrals and care-company acquisitions. Its team now exceeds 150 people, and the company reports service availability above 99%.
Limbic has solved a different set of problems. It connects with patient-management systems, collects clinical information, identifies risk, prepares assessments and supports patients between sessions.
Ash offers voice, text, memory, safety controls and a clinical advisory group. The consumer product is mature, although its healthcare infrastructure remains young.
The Path offers 11 AI therapist personalities, persistent memory, time-boxed sessions, crisis escalation and paid subscriptions. Independent treatment research and longer retention data are still missing.
Jimini’s Sage operates under a clinician’s direction and shares relevant information with the care team. Its design is clinically sensible, but deployment maturity remains difficult to judge without customer numbers.
| Startup | Product availability | Real-care integration | Current maturity |
|---|---|---|---|
| Wysa | Widely available across consumer and institutional channels | NHS, payors, employers, primary care and physical therapy | Most mature overall |
| Limbic | Deployed through healthcare providers | Deep NHS integration and expanding US presence | Most mature clinical system |
| Ash | Public consumer app using text and voice | Limited disclosed provider integration | Advanced consumer product |
| The Path | Public free and paid consumer product | No meaningful health-system scale disclosed | Fast-maturing consumer challenger |
| Jimini Health | Available through participating clinicians | Built specifically around professional oversight | Promising but early |
| Youper | Established consumer subscription app | Little recent institutional evidence | Mature self-help app |
| Earkick | Established consumer and workplace product | Limited disclosed clinical integration | Consumer-ready, clinically early |
| Sonia | Available voice and text product | No meaningful deployment data | Early commercial product |
Who is best prepared for tighter AI therapy rules?
Limbic is best prepared for tighter AI therapy rules today. Wysa comes second, and the consumer-first startups face a much harder adjustment.
Limbic Access holds UK Class IIa medical-device certification. That requires controlled development, documented risk management, clinical-safety processes and traceable testing. Limbic also works inside services where qualified professionals retain responsibility for treatment decisions.
Wysa’s AI provides structured support but does not diagnose users or make autonomous clinical decisions. Professional oversight remains available in its care programs, and referral information is passed to clinicians.
Ash has already felt the cost of regulatory uncertainty. Slingshot removed Ash from the United Kingdom after concluding that the country lacked a clear route for a wellbeing product that behaved increasingly like therapy.
The same pressure is growing in the United States. Illinois restricts AI from independently delivering therapy or making therapeutic decisions. Utah requires mental-health chatbots to disclose clearly that they are AI and places limits on how they use personal information. Nevada has also imposed restrictions.
The Path has built a separate crisis protocol that moves the AI away from normal conversation and toward human help when high risk is detected. Jimini’s clinician-supervised design also fits the direction regulators appear to prefer.
Youper, Earkick and Sonia disclose that they are AI and cannot handle emergencies. Those warnings may not be enough if regulators begin judging products by what they actually do rather than how they describe themselves.
If you want more recent data on this point, please see our latest mental health market report.
This chart, featured in our mental health market deck, illustrates yearly funding for mental health startups
Which AI therapy advantage is hardest to copy?
Limbic has the hardest workflow advantage to copy. Wysa has the deepest combination of distribution, research and accumulated usage.
A competitor could build a convincing mental-health chatbot within months using a strong open or commercial language model. Reproducing Limbic’s place inside NHS services would take much longer. The company has medical-device processes, EHR connections, clinical relationships and experience handling hundreds of thousands of assessments.
Wysa’s position comes from several advantages working together. It has years of conversation data, a large research program, millions of users and distribution through employers, payors, healthcare systems and public services. Its newer care businesses also give Wysa places to insert mental-health support without asking patients to download another app.
Ash’s model is more technically ambitious than the systems used by many older chatbot companies. Slingshot trained a foundation model around behavioral-health conversations and therapeutic methods instead of relying entirely on prompts wrapped around a general assistant.
That lead may shrink as OpenAI, Anthropic, Google and open-source developers improve memory, personalization, reasoning and safety. Ash still needs proprietary longitudinal data, proven outcomes and deeper distribution.
The Path’s adapted models, therapist personalities and Tony Robbins frameworks create a recognizable product, but its stronger defense would come from long-term outcomes and paid retention.
Is the best-funded AI therapy startup using its money efficiently?
Ash has the most money, but Wysa and Limbic have turned far less capital into far more proof.
Slingshot has raised $93 million, about three times Wysa’s reported total and more than six times Limbic’s. That capital paid for a purpose-built model, a large technical team, safety systems and free consumer distribution.
Ash has moved from a 50,000-person beta to roughly 200,000 users and generated a research dataset covering more than 100,000 accounts.
It has yet to show substantial revenue, named health-system deployments or a clear regulatory path in major markets outside the United States. Its own engagement study also found that more than half of accounts dropped out early.
As seen above, Wysa has reached more than six million users with roughly one-third of Ash’s funding. A crude funding-per-reported-user comparison produces around $5 for Wysa and $465 for Ash. This is not customer-acquisition cost: the companies are different ages, and Ash is funding a foundation model. Still, the 90-fold gap is striking.
Limbic has raised about $14.7 million while reaching close to half of NHS Talking Therapies services, completing more than 650,000 assessments and publishing several large clinical studies.
Jimini now has more capital than Limbic but remains early in deployment. Named customers and patient volumes will determine whether that funding was equally productive.
If you want more recent data on this point, please see our latest mental health market report.
This chart, featured in our mental health market deck, compares the main business model options for tele-mental health platforms
Which AI therapy startup has the strongest momentum now?
Limbic has the strongest momentum right now, with Ash and The Path moving fastest on the consumer side.
Limbic’s recent progress builds directly on its existing healthcare position. It published a major Nature Medicine study, expanded its clinical assessment volume and continued developing treatment-support and documentation products.
Ash has quadrupled its disclosed audience from the beta stage, launched voice conversations and published the first detailed analysis of how more than 100,000 people use a purpose-built generative mental-health system. The UK withdrawal slows its international path, but American growth remains strong.
The Path raised $14.3 million, passed 50,000 active members, recorded 2.5 million sessions and launched a $40 monthly plan. Its website also shows more than 2,700 reviews and a 4.9-star App Store rating.
Wysa’s recent moves are less visible but still meaningful. A £5.3 million Wellcome grant, the Kins acquisition and expansion through Singapore and underserved Indian communities all strengthen the business.
Jimini’s fresh $17 million round brings its total funding above $25 million. Its momentum currently comes from team-building and early provider partnerships rather than disclosed patient volumes.
Youper, Earkick and Sonia have released too little recent operating information to match this group.
Is AI therapy really one market?
AI therapy is already splitting into three markets, and no startup leads all three.
The first market is AI-guided entry into healthcare. These products collect information, assess needs, route patients and prepare clinicians. Limbic leads here, with Wysa also established.
The second market supports people who are already receiving care. Limbic Care and Jimini’s Sage help patients practise skills, complete therapeutic work and stay engaged between appointments. Human professionals remain part of the process.
The third market sells a direct relationship with an AI. Ash, The Path, Youper, Earkick and Sonia compete for people who want immediate, private conversations without arranging professional care. Product quality, memory, price and emotional connection matter more in this segment.
Wysa touches all three markets, which explains why it ranks first overall. It lacks Limbic’s depth in clinical intake and Ash’s generative-AI focus, but it can follow a person from anonymous self-help into coaching or formal care.
Consumer products will probably either stay firmly in wellness or invest heavily enough to enter regulated care. Clinical companies will keep AI under professional supervision while taking on more work between appointments.
This chart, featured in our mental health market deck, shows how market revenue is split across customer segments in the mental health market
How reliable are the AI therapy numbers?
The ranking is solid at the top and much less reliable below fourth place because these startups disclose different metrics and almost none reveal revenue or retention consistently.
Wysa’s numbers come mainly from the company. Six million users, one billion conversations and 11 million covered lives cannot be treated as equivalent measures. Covered lives describe eligibility through organizations, while conversations measure activity.
Limbic’s evidence is easier to interpret because many figures are tied to published NHS studies. Its current total of 650,000 assessments remains company-reported, while the earlier datasets of 129,400 and 64,862 patients appear in peer-reviewed journals.
Ash provides more engagement detail than most consumer competitors. Its 102,684-account study shows exactly how users were divided into behavioral groups. However, the company helped produce the research, and the clinical follow-up sample was tiny compared with the headline dataset.
The Path’s 50,000 active members and 2.5 million sessions describe current activity rather than lifetime downloads. The company has not disclosed the time window used to define an active member.
Funding numbers are easier to compare, but even those can vary because grants, accelerator money and unannounced rounds may be counted differently.
The weakest evidence sits with Sonia and Earkick. Their lower ranking reflects missing proof as much as weak products.
We have high confidence that Wysa and Limbic form the top two and that Ash belongs in third place. The order below Ash could change quickly.
Which AI therapy startups are actually ahead?
Wysa is ahead overall today, Limbic is the closest challenger, and Ash remains the startup most likely to change the order.
Wysa wins because it has already combined consumer reach, institutional distribution, research and operational maturity. None of its individual advantages is unbeatable. Together, they make it the only startup with a serious position across most of the category.
Limbic comes second by a smaller margin than the user counts suggest. It leads the most valuable clinical segment, has the strongest recent evidence and is better prepared for regulation. A successful expansion through large American providers could move it into first place.
Ash ranks third because it is the best-funded and fastest-growing generative-AI challenger. Its roughly 200,000 users remain far below Wysa’s reach, while more than half of the accounts in its engagement study left early. Better retention, controlled trials and paying demand would close the gap.
The Path moves directly into fourth place. Its current activity is already stronger than that of many older apps, while its VERA-MH result gives it an early safety argument. It still needs independent outcomes research and evidence that people will keep paying $40 per month.
Jimini ranks fifth on potential rather than present scale. Clinician supervision and large-provider distribution are sensible choices in today’s regulatory environment. Named customers and patient volumes could move it above The Path.
Youper remains sixth. Its three million historical users and published observational study give it more substance than most smaller apps, but the company has lost competitive momentum.
Earkick and Sonia complete the ranking. Both offer accessible products, though neither has disclosed enough current traction, clinical evidence or commercial progress to challenge the leaders.
Wysa holds a moderate overall lead. Limbic is almost level in clinical AI, while Ash is the strongest consumer challenger. The Path and Jimini are the two newer entrants most capable of disrupting the ranking.
| Rank | Startup | Why it ranks here |
|---|---|---|
| 1 | Wysa | The strongest overall mix of users, distribution, research, product breadth and operating maturity |
| 2 | Limbic | The clinical leader, with deep NHS adoption, medical-device certification and the best recent evidence |
| 3 | Slingshot AI, Ash | Rapid consumer growth and a purpose-built model, held back by retention, monetization and regulatory uncertainty |
| 4 | The Path | Strong early activity, a credible safety architecture and a live paid model, but limited independent outcome evidence |
| 5 | Jimini Health | A well-funded, clinician-supervised clinical challenger whose real deployment scale remains undisclosed |
| 6 | Youper | Large historical reach and useful research, with little evidence of strong current momentum |
| 7 | Earkick | An established and accessible self-care product with weak clinical and commercial disclosure |
| 8 | Sonia | A promising voice-based experience that remains too small and lightly validated to rank higher |
If you want more recent data on this point, please see our latest mental health market report.
This chart, featured in our mental health market deck, shows how therapy matchmaking platform technology has evolved over time
OUR METHODOLOGY
This analysis tests which independent AI therapy startup is actually ahead today. We compare eight companies whose AI speaks directly with people about mental health, either as a consumer product or as part of professional care.
We assessed reach, growth, customer value, clinical evidence, product maturity, regulatory readiness, defensibility, capital efficiency and recent momentum. No single metric settles the question: an app download, an active member, a covered life and a completed clinical assessment describe different kinds of progress.
We prioritized the freshest meaningful operating evidence available, including current usage disclosures, healthcare deployments, peer-reviewed studies, medical-device status, product expansion and signs of repeated adoption. Older figures were kept when they still provided the clearest scale benchmark.
We did not use a mechanical scorecard. The underlying evidence is too uneven, and forcing unlike metrics into one numerical model would create false precision. Instead, we judged what each figure actually demonstrates and looked for conclusions that held across several dimensions.
Funding was treated as capacity, not proof of leadership. Lifetime users show reach but not necessarily current engagement. Research volume shows commitment to evidence, but study design, sample quality and relevance to the current product determined how much weight each result received.
Where revenue and retention were not disclosed, we used named deployments, repeat activity, institutional adoption, workflow integration and measurable outcomes as the clearest available evidence of customer value. Company-reported figures were kept distinct from peer-reviewed results and regulatory records.
We excluded general-purpose assistants such as ChatGPT, Claude and Gemini, companion-first products such as Replika and Character.AI, and therapy marketplaces that still deliver care mainly through human clinicians. We also excluded companies whose current scale or evidence was too thin for a fair comparison.
The final ranking reflects the consistency of each company’s position across the full analysis. A startup did not need to lead every category, but it needed to convert its advantages into a credible, durable operating position rather than relying mainly on funding, technical novelty or a large historical user total.
Key sources used for this analysis include: Wysa’s company disclosures on reach, conversations and research, Wysa’s insurer and institutional distribution materials, Wysa Navigator and connected-care pathways, Limbic’s NHS adoption and assessment figures, the 2026 Nature Medicine study of Limbic’s cognitive-layer system, the Nature Medicine study of Limbic’s referral tool, the JMIR analysis of clinical efficiency and treatment pathways, Slingshot AI’s Ash launch and funding announcement, Ash’s engagement and outcomes preprint, The Path’s membership, session and product disclosures, Jimini Health’s latest funding announcement, Youper’s observational user study, the VERA-MH safety benchmark, Utah’s official mental-health chatbot framework, Nevada’s statutory restrictions, and the FDA’s description of the Breakthrough Devices Program.
In our mental health market deck, we identify pain points entrepreneurs should prioritize
