# HugoScore CAIHL Draft Report: Wysa

Status: Draft for human review
Last reviewed: 2026-06-10
Review method: Public-source review of the official website, clinical-evidence page, global privacy policy, FDA Breakthrough Device press release, peer-reviewed study records, NHS deployment pages, Mozilla privacy review, and merger press coverage; no app walkthrough, vendor interview, or independent model evaluation.
Service: Wysa
Vendor: Wysa (Touchkin eServices Pvt. Ltd. / Wysa Ltd.)
Category: Mental health AI

## Summary

Wysa is a CBT-based mental health chatbot with two faces. One is a free consumer app that anyone can use with only a nickname. The other is an institutional platform sold to employers, insurers, and health systems, including NHS Talking Therapies, where Wysa works as a digital front door, e-triage layer, waitlist support tool, and companion to human care. In March 2025 Wysa merged with April Health to add human collaborative care through primary care clinics, and it received a 2022 FDA Breakthrough Device designation for an AI conversational agent targeting chronic pain with depression and anxiety. The company claims more than 36 peer-reviewed publications and reports over 6 million users.

From a CAIHL perspective, the deployment channel changes the read. Used directly, Wysa is close to patient-directed: chosen by the user, anonymous by design, praised by Mozilla for privacy, and focused on self-help skills. Deployed institutionally, the same chatbot embeds organizational priorities. The institution decides that an AI conversation is the first offer, receives aggregate usage and safety reporting, and may be informed when the AI detects an emergency. Neither mode is automatically bad, but they should not be scored as one thing.

## Evidence Reviewed

- Wysa clinical evidence page: https://www.wysa.com/clinical-evidence
- Wysa global privacy policy, v7.2.0, revised 2026-06-05: https://legal.wysa.io/privacy-policy
- FDA Breakthrough Device designation press release, May 2022: https://www.businesswire.com/news/home/20220512005084/en/Wysa-Receives-FDA-Breakthrough-Device-Designation-for-AI-led-Mental-Health-Conversational-Agent
- Wysa and April Health merger press release, March 2025: https://www.businesswire.com/news/home/20250313775451/en/Wysa-and-April-Health-Merge-to-Revolutionize-Access-to-Behavioral-Health-Care-Through-Primary-Care-Providers
- Behavioral Health Business merger coverage: https://bhbusiness.com/2025/03/11/wysa-acquires-virtual-behavioral-health-provider-april-health/
- Wysa NHS Talking Therapies page: https://www.wysa.com/nhs-talking-therapies
- Wysa for Employers page: https://www.wysa.com/for-employers
- Wysa SOS crisis-detection study announcement: https://blogs.wysa.io/blog/company-news/ai-detects-82-of-mental-health-app-users-in-crisis-finds-wysas-global-study-released-on-the-role-of-ai-to-detect-and-manage-distress
- Role of AI in SOS page: https://www.wysa.com/role-of-ai-in-sos
- Mozilla Privacy Not Included review of Wysa: https://www.mozillafoundation.org/en/privacynotincluded/wysa/
- NHS Health Research Authority clinical investigation record: https://www.hra.nhs.uk/planning-and-improving-research/application-summaries/research-summaries/clinical-investigation-of-wysa-v10/
- 2018 chatbot safety investigation coverage (BBC reporting): https://geoffwhite.tech/2018/12/13/child-advice-chatbots-fail-to-spot-sexual-abuse/
- Imperial College 2026 project using Wysa in rural India: https://www.imperial.ac.uk/news/articles/2026/imperial-project-to-test-ai-mental-health-tool-for-girls-in-rural-india/

## CAIHL Profile

- Who does this AI serve? Patients in use, institutions in procurement. Individual users get free anonymous self-help, while employers, insurers, and health systems buy the platform and receive aggregate usage and configured safety reporting.
- Can patients tell AI is involved? Yes. Wysa is openly an AI chatbot, and the privacy policy states users are interacting with AI software and not a person.
- Can patients meaningfully choose? Yes in the consumer app, partial in institutional channels. In employer and NHS deployments Wysa is the offered first step, and declining it can mean waiting for human care.
- Can patients correct or challenge what the AI produces? Partial. Data correction, deletion, and reset-my-data rights are disclosed, but there is no public workflow for contesting an AI-supported triage outcome or an institutional report.
- Does it help patients understand or act? Yes. CBT tools, a clinician-written care library, SOS safety planning with helpline follow-up, and routes into human care are action-oriented, with peer-reviewed support in several populations.

## Agency Interpretation

Wysa's clearest agency value is low-barrier access to structured self-help. A person can practice CBT skills at 3 a.m. without giving a name, without cost, and with documented crisis signposting. The privacy design is unusually strong for the category: identifier scrubbing, zero data retention with the LLM provider, no conversation training by the external LLM, and a real deletion control. Mozilla rated it among the best mental health apps it reviewed.

The unresolved CAIHL tension is the institutional channel. When an employer or health system deploys Wysa as the front door to care, the patient did not choose the AI, the institution did. Aggregate dashboards define value in workforce and utilization terms, safety detections may be reported to the institution, and the chatbot can function as a buffer before scarce human therapy. That is exactly the configuration CAIHL says to watch, even when the underlying product is well-built. The 2018 BBC-reported failures on child abuse disclosures are dated and were reportedly fixed, but they remain a reminder that crisis detection claims need independent verification.

## Key Unknowns

- Whether any FDA marketing authorization has followed the 2022 Breakthrough Device designation. None was found in this review.
- Exactly what safety and usage data each institutional deployment shares with the employer, insurer, or NHS service, and what users are told at signup.
- How the post-merger April Health collaborative care model changes data flows between the chatbot, clinicians, and primary care billing.
- Independent verification of SOS crisis-detection performance beyond the company's own 82 percent detection study.
- Whether users in NHS e-triage can contest or bypass AI-supported routing decisions.
- Current effectiveness evidence for the LLM-enabled conversation mode specifically, as distinct from the older rules-based product that most published studies evaluated.

## Publication Recommendation

Ready for human review as a draft profile. Confidence should stay at medium until the FDA status is checked against FDA databases directly, institutional data-sharing configurations are verified, and at least one independent evaluation of the current LLM-enabled product is found.