Innovation & Tech

Apple Taps OpenEvidence for a 200,000-Person Study of Health AI

Apple and OpenEvidence will enrol up to 200,000 US iPhone users in a 12-month study of AI health answers grounded in NEJM, JAMA and Nature, with queries deleted after 24 hours.

By Nathan Brooks

3 min read

Apple Partners With OpenEvidence on Large-Scale Study of Health AI Use
Apple Partners With OpenEvidence on Large-Scale Study of Health AI Usetiburi / pixabay

What's News

  • Apple and OpenEvidence will enrol up to 200,000 participants in a study running more than 12 months.
  • The app is open to US users over 21 with an iPhone 15 or later, iOS 26 or later and Apple Intelligence enabled.
  • OpenEvidence deletes queries and responses after 24 hours, and data will not be used to train AI models.
  • Answers draw on the New England Journal of Medicine, JAMA and Nature.
  • Users can share HealthKit data including resting heart rate, activity, sleep and lab results.

Apple will enrol up to 200,000 people in a year-long study of how consumers actually use AI for health questions, partnering with healthcare AI company OpenEvidence on a research app that is available to download now.

The study runs for more than 12 months and is open to US users over 21 who own an iPhone 15 or later, run iOS 26 or later, and have Apple Intelligence enabled. That hardware floor cuts off a meaningful chunk of the installed base — anyone on an iPhone 14 or earlier simply can't participate, which matters when you're trying to draw population-level conclusions.

Participants ask health questions and receive simplified answers grounded in clinical sources, including The New England Journal of Medicine, the Journal of the American Medical Association and Nature. The app goes beyond a chatbot format in several ways:

  • Users can rate responses and give feedback on answer quality
  • They complete monthly surveys tracking outcomes over time
  • They may share HealthKit data — resting heart rate, activity, sleep and lab results — for more personalised guidance
  • The app helps users prepare for doctor's visits and directs them to providers when answers raise concerning issues

That last point is doing quiet strategic work. Apple is positioning the tool as a bridge to clinicians, not a replacement, and the referral mechanism is baked into the product rather than left to user judgment.

How does Apple handle the data?

This is the question practitioners and privacy watchers will ask first, and Apple answered it preemptively. The company said all data will be anonymised before it reaches a secure third-party cloud. It will not be used to train AI models. OpenEvidence deletes queries and responses after 24 hours, which means the raw interaction record is short-lived by design.

For context, that 24-hour deletion window is far stricter than typical consumer AI products, where queries often persist for months to improve models. Apple is explicitly ruling out that use case — a notable constraint given how hungry AI developers usually are for real-world health query data.

What does this change for health AI?

Sumbul Desai, Apple's vice president of Health and Fitness, framed the stakes in a statement: responsible innovation requires scientific rigour and clinical grounding, and the study will help build trusted tools that support people's health.

The study's real contribution is evidence. Most consumer health AI tools ship without published research on whether their answers genuinely help users — this one is designed to generate exactly that dataset, at a scale of up to 200,000 participants that few academic trials reach.

For clinicians, the likely near-term effect is patient arrival patterns: more people showing up to appointments having asked an AI grounded in NEJM and JAMA literature, possibly with HealthKit data in hand. For Apple's competitors, it raises the bar on transparency — the privacy commitments, journal-grounded sourcing and clinical referral flow are now the benchmark other health AI products will be measured against.

If the study shows clinically grounded answers improve user outcomes, expect Apple to fold the findings into its broader health stack. If it doesn't, the results will be just as instructive for an industry still guessing at what health AI users actually need.

Source: theaiinsider.tech

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.

2,251 articles

Related articles

« Previous articleNext article »