On a quiet Sunday, Apple opened a research study that lets people ask everyday health questions and receive answers it says rest only on peer-reviewed medicine, with the option of tailoring those answers to their own phone data. It arrives at a moment when millions have already turned to AI chatbots for guidance about their bodies, often with mixed results, and when the distance between an accurate answer and a useful one remains wide. The real question is not whether the machine knows the medicine, but whether ordinary people can carry that knowledge into their lives and into conversations wi
Apple Launches AI Health Study App Built on Peer-Reviewed Research and Opt-In Health Data
A bridge, not a barrier, to better care
Why does this launch matter, when chatbots already answer health questions?
Because Apple is framing it as a study, not just a product. It wants to measure whether people trust the answers, understand them, and use them to talk better with their doctors.
The Oxford numbers seem central to that.
They are. Chatbots identified conditions nearly 95% of the time in clinical cases, but everyday users got to the right answer under 35% of the time. The gap is in understanding, not just accuracy.
Worth saying Apple hasn't shown its tool closes that gap. The study is designed to find that out, so nothing here is a result yet.
And the privacy claims?
Apple says identifying information is removed before questions reach OpenEvidence, which deletes data within 24 hours and doesn't train on queries.
Those are Apple's statements as relayed to ABC. The reporting doesn't describe any independent verification of them.
What about the safety safeguards?
The app is meant to recognize dangerous situations and point people to professional care, and it says it isn't live monitoring or a replacement for a clinician.
How well it recognizes those situations is unknown. We're told the safeguards exist, not how they perform.
Who can actually take part?
U.S. adults 21 and older with an iPhone 15 Pro or later on iOS 26 or later, after an informed consent process.
Der Puls
- A widely used technology is entering the most personal corner of daily life, where a wrong answer can mean false comfort or a delayed visit.
- Doctors and researchers warn that accurate information is not automatically useful, and that confusing guidance can quietly do harm.
- Apple is working to ground answers in peer-reviewed citations, strip identifying details from questions, and tie advice to each user's own health metrics only if they opt in.
- The study is now enrolling eligible U.S. adults, and its findings on whether people truly understand and trust the answers will decide whether this bridge holds.
On a quiet Sunday, Apple opened a research study that lets people ask everyday health questions and receive answers it says rest only on peer-reviewed medicine, with the option of tailoring those answers to their own phone data. It arrives at a moment when millions have already turned to AI chatbots for guidance about their bodies, often with mixed results, and when the distance between an accurate answer and a useful one remains wide. The real question is not whether the machine knows the medicine, but whether ordinary people can carry that knowledge into their lives and into conversations with their doctors without being misled. Apple has chosen to find out in the open, measuring trust and understanding before it claims success.
On Sunday, Apple released the Apple Health AI Usage Study, a research app that answers everyday health questions using what the company describes as trusted, peer-reviewed medical research. Users who choose to share data already in their phone's health app, such as resting heart rate, sleep, or blood test results, can receive answers tailored to their own numbers rather than generic advice.
The launch follows a surge of people turning to AI chatbots for health information, a trend accompanied by troubling reports of wrong or harmful answers. Apple's approach relies on a partnership with OpenEvidence, a tool used by doctors and trained on medical journals that many other platforms cannot access. Each answer links to clinical citations so users can trace where the information came from.
Privacy is central to the design. Apple says questions are stripped of identifying details before reaching OpenEvidence, which deletes them within 24 hours and does not use them for training. The app is built to recognize dangerous situations and point people toward professional help, and it states plainly that it offers no live medical monitoring and does not replace a clinician.
The study itself will measure trust, comprehension, helpfulness, and whether the tool helps people prepare for doctor visits. Dr. Sumbul Desai of Apple said the goal is for AI to serve as a bridge to care. Harvard's Dr. John Brownstein cautioned that accuracy alone is not enough, and that turning personal health data into something people can act on remains difficult. Eligible participants are U.S. adults aged 21 and older with an iPhone 15 Pro or later running iOS 26.
On Sunday, Apple released a new app called the Apple Health AI Usage Study, a research project that lets people ask everyday health questions and get answers the company says are drawn only from trusted medical research. The answers can also be tailored to a user's own health metrics and records, but only if the user chooses to allow that.
The launch comes as AI chatbots become a common place to go for health information, alongside a steady stream of headlines about answers that are wrong or even harmful. According to the Pew Research Center, 35% of adults, and about half of adults under 30, already use AI chatbots. A February 2026 Oxford study published in Nature Medicine found that chatbots correctly identify medical conditions nearly 95% of the time when tested on clinical cases. Ordinary users trying to make sense of dense guidance reached the right answer less than 35% of the time.
Apple's approach is built on a partnership with OpenEvidence, an AI tool used by doctors and trained on sources such as medical journals that may not be available to other AI platforms. Full access to OpenEvidence is currently limited to health care professionals, and it returns answers in language they are used to. Apple says it is using its own technology to turn that research into plain-language answers. Each response comes with linked, peer-reviewed clinical citations so users can see where the information originates.
People who join can also share data already stored in their phone's health app, including resting heart rate, activity, sleep or blood test results. Someone asking how to improve their resting heart rate, for example, could get advice based on their own measurements instead of generic recommendations. Sharing is optional.
The study itself is meant to measure how people experience all this: how much they trust the information, how well they understand it, how helpful it is, and whether it helps them interact with their doctors. Participants can rate answers and complete occasional surveys. The study will also look at whether the tool helps people prepare for visits with their own clinicians.
Dr. Sumbul Desai, Apple's vice president of Health and Fitness, told ABC News that interactions need to be responsible, easy to understand and able to help someone navigate traditional care. She said people with clear, medically grounded information can walk into appointments feeling more prepared, and that rigorous study is needed so AI serves as a bridge rather than a barrier to better care. Dr. John Brownstein, a Harvard Medical School professor of pediatrics and biomedical informatics and an ABC News contributor, said accurate answers are not always useful, and that confusing or incomplete information can create false reassurance or delay care. He also noted that phones and wearables collect enormous amounts of personal health data, but turning it into something people can act on remains a challenge.
On privacy, Apple says questions pass through protections designed to strip identifying information before they reach OpenEvidence. According to Apple, OpenEvidence will delete question-and-response data within 24 hours and will not train its model on the queries. The app includes safeguards meant to recognize potentially dangerous situations and steer users toward professional help. Apple stresses that the study does not provide live medical monitoring or replace a health care professional, and says the app will state this while in use.
The study is open to U.S. adults 21 and older with an Apple Intelligence-enabled iPhone, meaning an iPhone 15 Pro or later, running iOS 26 or later. Interested people download the dedicated study app, learn about the research and complete an informed consent process before enrolling. Apple told ABC News it intends to use the findings to inform future products and also to make the research broadly available. What the results show about whether people understand and appropriately trust these answers is what to watch.
Bemerkenswerte Zitate
Rigorously studying these experiences is essential to ensuring AI serves as a bridge, not a barrier, to better care.— Dr. Sumbul Desai, Apple vice president of Health and Fitness
We need to understand not just whether AI provides accurate answers, but whether people understand them, trust them appropriately and make better health decisions.— Dr. John Brownstein, Harvard Medical School and ABC News contributor