Apple recruits up to 200,000 people to test health AI

Apple has launched the Apple Health AI Usage Study in the United States, an online health AI study built around a dedicated research app. Its partner is OpenEvidence, a clinical AI platform. The study aims to answer four questions: whether users trust conversational health AI, whether the AI's answers are good, whether users can understand them, and whether their everyday health behavior changes afterward.

The eligibility requirements are specific. Apple will enroll up to 200,000 people who are at least 21, live in the US, and own an iPhone that supports Apple Intelligence, meaning an iPhone 15 Pro or later running iOS 26 or newer. Participants download the research app and sign up voluntarily.

Illustration: clay figures ask health questions on phones; tiny books symbolize medical AI answers citing literature

What the research app does

Inside the app, participants can ask everyday health questions. The AI explains complex medical information in plain language, and every answer comes with clickable citations to peer-reviewed literature. OpenEvidence is already widely used by physicians as a clinical AI tool, and its answers are generated from sources such as the New England Journal of Medicine, JAMA and Nature.

Participants can also choose to share data from HealthKit, such as resting heart rate, activity levels, sleep and lab results. The researchers want to see whether answers become more personal and easier to understand once personal data is connected. Apple also wants to know whether health AI can help people sort out their questions before they see a doctor.

On privacy and safety, Apple has made several explicit commitments: study data will not be used to train AI models; participants decide for themselves whether to share HealthKit data and questionnaire responses; and the app has built-in safeguards that suggest seeing a doctor or using other resources when it detects certain concerning questions. Apple also states that the app does not diagnose or treat conditions and cannot replace a doctor's advice.

“Patients and clinicians are both turning to AI, but the tools they use are being built and tested in separate worlds.”

The line comes from Robert Wachter, chair of the Department of Medicine at the University of California, San Francisco. Sumbul Desai, Apple's vice president of Health and Fitness, said in the announcement that health is deeply personal, and that a dedicated study is the most careful way to understand how people use and evaluate health AI.

Why Apple is not simply shipping a feature

Quite a few companies have already built conversational health Q&A into their products, while Apple chose to run a study first. It has run large online studies in health before, such as its heart and hearing studies, and this one follows the same playbook. With large language models in the mix, the hard part shifts from sensor accuracy to answer quality. Apple itself acknowledges in the announcement that consumer health tools built on large language models can sometimes give incomplete or vague answers.

The announcement does not give a study duration, nor does it say when or in what form results will be published. Apple has also not said whether, or how, the user behavior data gathered in the study might eventually be used in its own Health app.

Users in mainland China cannot take part for now

The study is open only in the US, so users in mainland China cannot sign up. Apple Intelligence on mainland China devices is still in testing, with Alibaba's Qwen as the partner. Reports on the first batch of features point to general capabilities such as notification summaries and Genmoji, and health Q&A is not among them.

OpenEvidence relies on the English-language medical journal system. Moving to a Chinese-language setting would mean starting over on literature sources, prescribing habits and insurance rules. Phone makers in China that want to build similar health AI also face different partners and different clinical data interfaces than in the US. In the short term, the study matters to Chinese-speaking users mainly as a method: a major company is willing to put in the effort to test whether people trust and understand the AI before deciding whether to roll the feature out to everyone.

Sources: Apple Newsroom announcement, CocoLoop, OpenEvidence, ITHome; the enrollment cap, age requirement and device and OS requirements follow Apple's announcement.