Apple tests an AI assistant to better understand connected health
Apple launches a study in the United States on the use of AI-based health tools, in partnership with OpenEvidence. Up to 200,000 people can participate via an iPhone app, with a strong focus on privacy.
Apple has launched the Apple Health AI Usage Study in the United States, a study designed to measure how users interact with conversational health tools and how they assess their responses. Conducted with OpenEvidence, a clinical AI platform used by professionals, it relies on an iPhone app where participants ask medical questions, evaluate the clarity of the answers, and, if they wish, share HealthKit data. The challenge goes beyond a simple product test: it is about determining whether a health AI can be useful, understandable, and trustworthy without crossing the line into diagnosis.
Une étude de confiance, pas un lancement de produit médical
Apple presents this initiative as a research project, not as a diagnostic service. The company wants to observe the quality of the answers, but also how people understand these answers and use them in their daily lives, notably to prepare for a consultation or better interpret a symptom. The distinction matters because health tools based on large language models can produce incomplete or too vague explanations; Apple is therefore trying to measure what works and what doesn’t.
OpenEvidence provides the clinical basis for the experience, with answers supported by medical sources such as The New England Journal of Medicine, the Journal of the American Medical Association, or Nature. Participants receive explanations in simple language, accompanied by verifiable clinical citations. This is a way to remind that, in health, AI must learn to speak clearly without playing the amateur doctor.
Une app iPhone, des questions libres et des retours notés
The application is available for download and registration began on November 15, 2026, according to information provided by Apple. The program can accommodate up to 200,000 people, provided they are 21 years or older, reside in the United States, and use an iPhone 15 Pro or newer running iOS 26 or later. The app does not use Siri or Apple Intelligence, although it accepts text input and, via the iOS keyboard, dictation.
The operation is simple: users ask a health question, read the answer, then rate it according to several criteria, including accuracy, level of detail, and relevance. Apple thus wants to understand if an answer is not only correct in substance but truly helps to act, prepare a medical appointment, or better situate a health problem in a personal context.
La donnée personnelle, mais sous contrôle
Personalization is optional. During registration, participants can authorize the sharing of HealthKit data such as activity, sleep, resting heart rate, or lab results. The idea is to verify if this context improves the relevance of the answers, but Apple emphasizes a strengthened privacy framework and states that the collected data will not be used to train its AI models.
This promise does not remove all nuances. The App Store privacy sheet indicates that user identifiers may be collected and linked to identity for analysis and app operation purposes, while diagnostic and performance data may also be recorded. In short, the study explores the boundary between useful personalization and measured data collection, without claiming everything is black or white [to be verified].
Ce que cette étude dit du marché de la santé IA
Apple is not only trying to improve an app. The company is testing a broader question: how can conversational health tools become credible for general public use without blurring the line between information and care? Robert M. Wachter, professor and chair of the UCSF Department of Medicine, summarized the challenge by saying: “When patients and doctors draw on versions of the same trustworthy tools, patients arrive prepared, conversations improve, and next steps happen seamlessly,” which well expresses the ambition to bring patient tools closer to those of clinicians.
In this context, Apple follows a cautious path. The study provides neither medical advice, diagnosis, nor treatment. It focuses mainly on trust, understanding, and practical uses. It is probably slower than a flashy launch but likely more useful in the long run for a sector where the slightest ambiguous answer can have very real consequences.
Key points
- Apple launches an AI health study in the United States on November 15, 2026.
- Up to 200,000 people can participate via an iPhone app.
- Participants must be 21 years or older and have an iPhone 15 Pro.
- OpenEvidence provides clinical answers based on medical sources.
- Apple states it does not use this data to train its AI models.
- The research measures trust, understanding, and quality of responses.
By the numbers
- 200,000 people — maximum announced capacity, United States, Apple.
- 21 years — minimum age required to participate, United States, Apple.
- iPhone 15 Pro — minimum required model, iOS 26 or later, Apple.
- November 15, 2026 — launch date announced by Apple.