Apple Launches Apple Health AI Study App for Up to 200,000 Users
Apple's new app recruits 200,000 US adults for a 12-month study of how people understand, trust, and act on health AI answers, powered by OpenEvidence.

Why it matters
- Apple launched the Apple Health AI Study app on Sunday, open to 200,000 US participants aged 21+.
- Participation requires an Apple Intelligence-enabled iPhone (iPhone 15 Pro or later) running iOS 26 or later.
- Apple's partner OpenEvidence draws on NEJM, JAMA, NCCN, Nature and Cochrane, and deletes query data within 24 hours.
- Study data will not be used to train AI models but may inform Apple's feature development.
- One in three US adults already use AI chatbots for health information, per figures Apple cited.
Apple launched a new research app on Sunday that will recruit up to 200,000 US participants to study how people actually use conversational AI for health questions — a population that already includes one in three US adults, according to figures Apple cited during a press briefing.
The Apple Health AI Study continues a line of research Apple has run since 2019 through its Research app, covering hearing, heart health, movement, and women's health. This time the subject is not a biomedical condition but the behavior of people talking to AI about their health. The stakes are straightforward: health chatbots are already widely used, and nobody — including Apple — has settled evidence on whether users understand the answers, trust them, or act on them.
What will the study actually measure?
Apple said the study will examine four things: comprehension, response quality, the drivers of user trust, and everyday health actions. The company frames the goal as informing "the responsible design of health AI experiences for both Apple and the industry" and exploring how health AI might help people prepare for doctor's visits.
The key questions Apple says it wants answered are basic and, for a company building health features, uncomfortable to leave open: Do users understand the information provided by health AI? Do they trust it? Is it helpful?
Apple stated that study data will not be used to train AI models. It may, however, inform Apple's feature development — which means the research doubles as product input, a caveat worth keeping in mind when reading the eventual results.
The study will also collect demographic information — race, age, education level — to explore how those factors affect preferences and communication needs. That is a rare attempt to disaggregate health AI behavior by population, and it addresses a real gap in how generic chatbot benchmarks treat users as interchangeable.
Who is OpenEvidence, and what do participants get?
Apple's study partner is OpenEvidence, an AI platform built specifically for doctors. It generates answers using medical information from a defined set of sources: the New England Journal of Medicine, the Journal of the American Medical Association, the National Comprehensive Cancer Network, Nature, and the Cochrane Database of Systematic Reviews.
Inside the dedicated study app, participants type health questions and receive answers with linked, peer-reviewed clinical citations. They can also opt to share HealthKit data — sleep, lab results, activity, and resting heart rate — for more personalized answers. HealthKit is Apple's hub for health and fitness data collected on the iPhone and Apple Watch.
Questions can cover general health, specific symptoms, recovery from a procedure, and the participant's own uploaded health data. That citation-backed design distinguishes the study from consumer chatbots that pull from general web corpora, and it means the findings may not transfer neatly to tools without clinical sourcing.
Who can join, and what hardware is required?
The eligibility list is narrow, and the hardware requirement is the tightest constraint:
- US residency
- Age 21 or older
- An Apple Intelligence-enabled iPhone — iPhone 15 Pro or later
- iOS 26 or later
Apple says the study app requires the processing capabilities of those iPhones because it uses on-device Apple Foundation Models, the large language models that are part of Apple Intelligence. Those models provide access to any large language model and, within the study, are used to classify question categories — meaning the phone itself sorts what a question is about before anything leaves the device.
What does participation involve?
Each participant commits for up to 12 months. The workload includes:
- A baseline survey
- Monthly surveys
- Additional surveys specific to preparing for doctor's visits
- Periodic check-in prompts, including Apple Watch notifications with suggested topics to explore
Apple notes that a complete lack of participation may result in removal from the study — a reasonable retention mechanism, though it also selects for engaged users, which could skew results toward people already inclined to trust health AI.
How does Apple handle the data?
The privacy pipeline is the most technically detailed part of the announcement, and Apple laid it out in steps. After a mandatory onboarding process that includes informed consent, the iPhone first processes questions using tools designed to remove details that could identify participants. Queries then go to a secure cloud-based study environment managed by a third-party contractor, UST Global, where they are delinked from individual participant IDs and screened again for remaining personal details. Only then is the information sent to OpenEvidence, where answers are generated with supporting reference materials.
OpenEvidence will delete both question and response data within 24 hours, Apple said.
Access on Apple's side is restricted: a small group of Apple researchers will see coded versions of survey responses separated from personally identifiable information, summaries of conversations, and categories of the most common conversation types.
HealthKit sharing is strictly opt-in and double-gated. Participants must first choose to upload clinical health records to Apple's Health app, then separately choose to share those records from the Health app into the Apple Health AI Study app.
What are the safety guardrails — and their limits?
The study app includes guardrails that detect concerning questions and respond by suggesting the participant follow up with a provider or other resources. But Apple was explicit about the limits: there is no live monitoring, and study staff will not contact participants even if questions related to self-harm are asked.
That decision is a defensible privacy trade-off — monitoring would require exactly the identification the pipeline works to strip out — but it means the guardrail is a suggestion, not an intervention. Participants in distress get routed to resources by software, not people.
Why this matters now
One in three US adults already use AI chatbots for health information, and Apple is positioning itself to shape how the industry builds for them. A study of 200,000 people examining comprehension, trust, and behavior at that scale would be among the largest systematic looks at health AI usage to date — and Apple controls the design, the partner, and the eventual publication.
The Apple Health AI Study app is available now in the App Store. Apple has set no publication date for the study's findings, so the gap between data collection and public results — and how much of it shapes Apple's own products before anyone outside sees it — remains an open question.
Source: cnet.com
More from James Calloway
Show full bio
News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.
610 articles
Related articles
- OpenAI Launches Health in ChatGPT With Medical Record Connections
- Oura Will Pay $625 Million for AI Health Startup Counsel Health
- OpenAI Releases MentalHealthBench, an Expert-Built AI Mental Health Benchmark
- OpenAI Launches ChatGPT Health, a Dedicated Health Data Experience
- Google Releases First Validated Toolkit for Measuring AI Manipulation