Monday, 16 November 2026

Apple opens a health AI study app to 200,000 iPhone owners

Apple's Health AI Usage Study, built with OpenEvidence, lets up to 200,000 US iPhone owners ask health questions and answer surveys about the replies.

Key points

  • Apple's study app is open to up to 200,000 US participants aged 21 and over.
  • It needs an iPhone 15 Pro or newer on iOS 26 or later.
  • Answers come from OpenEvidence with peer-reviewed citations; study data won't train AI models.

Apple has launched a research app that lets iPhone owners put everyday health questions to an AI chatbot, then judge how clear and trustworthy the answers were. The Apple Health AI Usage Study runs virtually, and Apple says it will enrol up to 200,000 people.

The catch for most of our readers is the location: only adults aged 21 and over in the United States can sign up. The phone matters too, because the app needs an iPhone 15 Pro or newer with Apple Intelligence switched on, on iOS 26 or later.

The chatbot side is not Apple’s own. It comes from OpenEvidence, a clinical AI service used by doctors and other healthcare staff, which draws on journals including NEJM, JAMA and Nature. Every reply carries linked, peer-reviewed citations, so a participant can open the paper behind a claim. Apple built the app around it, with starter prompts such as summarising your vitals, sleeping better or getting ready for a doctor’s visit.

The app can read HealthKit too, if you allow it: pulse at rest, daily activity, sleep, even results from lab tests. The point, per Apple’s announcement, is to see whether personal data makes answers more relevant. Nothing is compulsory, neither the HealthKit sharing nor the survey answers, and Apple says none of it will go into training AI models.

What Apple is measuring is trust, comprehension and what people actually do after reading the advice. It says the findings should shape how health AI is designed, at Apple and elsewhere. One strand looks at whether AI can help patients prepare for appointments, and the app can produce a visit-prep summary for that.

Apple also brought in an outside voice. Robert Wachter, who chairs the Department of Medicine at UCSF, argues that patients and doctors now use AI tools built and tested separately. His view is that when both rely on the same kind of trustworthy tool, patients turn up better prepared and the follow-up goes more smoothly. That is the bridge the visit-prep feature is meant to test, with the summary a participant could carry into a real consultation.

Inside the app, participants also answer survey questions asking them to rate the quality of what the chatbot told them, which is how Apple gathers the trust and comprehension data. Apple has not published results, a timeline or what it plans to build from them.

Apple concedes the weakness it is probing: tools built on large language models sometimes give incomplete or unclear health answers. The app has guardrails meant to catch worrying questions and steer people to a doctor or other help, and Apple stresses that it cannot diagnose or treat anything. Sumbul Desai, who leads Health and Fitness at Apple, said “responsible innovation starts with scientific rigor and clinical grounding.”

Maroof Saeed

Author/Writer at Tech Prolonged You will mostly see him posting about Android stuff. Reach him on social networks below. @Twitter Facebook