Apple Recruits 200,000 Americans to Test Health AI With OpenEvidence
Apple is teaming up with clinical AI platform OpenEvidence to study how people actually use AI for health questions, with up to 200,000 US participants and optional HealthKit data sharing. It lands as surveys show roughly a third of US adults already ask chatbots about their health, while research flags shaky accuracy and blind trust in the answers.
- Original source: gizbot.com.
- The source published it on Nov 16, 2026 at 10:35 AM.
- No second outlet has reported it yet. Treat it as a single-source claim.
Apple's newly announced Apple Health AI Usage Study will enroll as many as 200,000 people in the US to ask conversational AI everyday health questions and see whether they trust what comes back. The company is running it with OpenEvidence, a clinical AI platform used by healthcare professionals whose answers draw on established medical publications including The New England Journal of Medicine, the Journal of the American Medical Association and Nature. Participants get plain-language explanations of complicated medical topics plus links to the peer-reviewed clinical research behind each response.
Participants can also opt to share Apple Health data through HealthKit, covering resting heart rate, activity levels, sleep information and lab results. Apple wants to learn whether handing AI that personal context makes its answers more relevant and easier to follow. The study additionally looks at whether AI can help people prepare for doctor's appointments and hold more informed conversations about their health.
The timing is not subtle. A March 2026 KFF survey found 32% of US adults had used AI chatbots for health information or advice in the past year, and about 19% had leaned on them to understand medical tests, lab results or diagnoses. That is no longer a niche habit.
Accuracy is the sore spot. A November 2024 study in the Journal of Medical Internet Research scored four AI chatbots on emergency medical care questions: 85% for clarity and understandability, but only around 50% for both accuracy and completeness, with between 5% and 35% of responses containing potentially dangerous information. A separate 2026 JMIR study ran two experiments with 338 college students and 563 participants recruited via Amazon Mechanical Turk and found that people who depend more on AI are less sensitive to inaccuracies, making them likelier to swallow wrong answers.
On privacy, Apple says study data will not be used to train AI models, and participants keep control over whether they share HealthKit metrics or survey responses. The study app includes safety measures to spot certain concerning questions and point people to a healthcare provider or other resources. Enrollment is capped at 200,000 and open to US residents aged 21 and up with an iPhone 15 Pro or newer running iOS 26 or later with Apple Intelligence support, via the Apple Health AI Study app. Apple notes the app is not built to diagnose or treat conditions or to replace professional medical advice.
Apple already has a deep health footprint through the iPhone and Apple Watch, and folding AI into that ecosystem could eventually help users read their own health data and walk into appointments better prepared. Hitting that bar means a level of accuracy and trust that general-purpose chatbots have not reliably delivered.
TopicsApple · OpenEvidence · Apple Health · HealthKit · Apple Health AI Usage Study · KFF · Journal of Medical Internet Research · Apple Intelligence
Source: gizbot.com
Written by AI from verified reporting.

