Apple Launches New App to Study How People Use AI for Health - CNET
Apple has launched a 12-month study app that asks 200,000 US adults aged 21 and up to use a health chatbot built on OpenEvidence, which answers from peer-reviewed sources like NEJM and JAMA. The study is measuring whether people understand, trust, and find useful the answers health AI gives them.
- Participants need an iPhone 15 Pro or later running iOS 26, and can optionally share HealthKit data such as sleep, labs, and heart rate for more personal answers.
- Questions are stripped of identifying details on the phone, routed through a third-party contractor, then sent to OpenEvidence, which deletes question and response data within 24 hours.
- Study data won't be used to train AI models, though it may shape Apple's own health feature development. There's no live monitoring, and self-harm questions trigger only an automated suggestion to seek help.
Apple has launched the Apple Health AI Study, a 12-month research project built around an app that lets participants put health questions to a conversational AI and see answers backed by peer-reviewed citations. The study is open to 200,000 US participants aged 21 and older, and they need an iPhone 15 Pro or later running iOS 26 with Apple Intelligence. Apple says it wants to learn how people understand health AI answers, how useful they find them, what makes them trust them, and whether the tools help people prepare for doctor visits. The timing matters because, according to the article, one in three US adults already use AI chatbots for health information, and Apple is trying to gather evidence on whether those answers hold up.
The AI itself is built by OpenEvidence, a platform aimed at doctors that draws on sources including the New England Journal of Medicine, JAMA, Nature, the National Comprehensive Cancer Network and the Cochrane Database of Systematic Reviews. Participants can optionally share HealthKit data such as sleep, lab results, activity and resting heart rate, and Apple's own Health app records can be included if they choose to upload them first. Participation involves a baseline survey, monthly surveys, extra surveys about doctor visits, and periodic prompts from the app, with notifications on paired Apple Watches. Apple says study data won't train its AI models, though it may shape future feature development.
On privacy, the chain is layered. Questions are stripped of identifying details on the iPhone, then sent to a cloud environment run by a third-party contractor, UST Global, where they are detached from participant IDs and screened again. OpenEvidence then generates answers and deletes both questions and responses within 24 hours. Only a small group of Apple researchers sees coded survey data and summaries of conversation categories. The app has guardrails that flag concerning questions and suggest contacting a provider, but there is no live monitoring, and study staff won't reach out to people who ask about self-harm. There is also no set publication date for the study's findings.