On November 15, 2026, Apple launched the Apple Health AI Usage Study, a virtual research study with its own iPhone app, open to adults in the United States. Built with OpenEvidence, a clinical AI platform used by healthcare professionals, the app answers everyday health questions in plain language and attaches linked, peer-reviewed clinical citations to each response. Apple says it wants to understand what drives user trust, response quality, comprehension and everyday health actions, and to use what it learns to inform the responsible design of health AI for itself and the industry.
Apple aims to enroll up to 200,000 participants. According to TechCrunch, the study will run for more than 12 months.
/ 01What the study app does
Participants ask health questions and get answers grounded in clinical sources such as The New England Journal of Medicine, the Journal of the American Medical Association and Nature. They can explore complex medical information in simpler terms, rate the quality of responses, and fill in surveys; TechCrunch reports the surveys are monthly. The app is also meant to help people prepare for visits with their own doctors.
Participants can choose to share HealthKit data, such as resting heart rate, activity, sleep or lab results, so the study can see how personal data affects the relevance and understanding of answers. Apple says the app has safety guardrails that detect certain concerning questions and suggest following up with a provider or other resources, and that it is not meant to diagnose or treat any condition or replace medical advice.
Eligibility, per Apple’s announcement: age 21 or older, living in the US, with an Apple Intelligence-enabled iPhone (iPhone 15 Pro or later) running iOS 26 or later.
On data, Apple says study data will not be used to train AI models and that participants control whether they share HealthKit metrics or survey responses. TechCrunch adds that data is anonymized before being sent to a secure third-party cloud provider, and that OpenEvidence deletes query and response data after 24 hours.
/ 02Why it matters
People already use general-purpose chatbots for medical questions at scale. TechCrunch notes that OpenAI said earlier this year users were submitting more than 300 million health-related queries a week, and that ChatGPT Health was made available to all US users in July. Apple is taking a different route: instead of shipping a health assistant, it is running a study to measure whether clinically grounded answers actually help.
Robert M. Wachter, chair of the Department of Medicine at UCSF, put the gap this way in Apple’s release: patients and clinicians are “both turning to AI,” but their tools are built and tested separately. His argument is that shared, trustworthy tools would help patients arrive prepared and improve conversations with doctors. Sumbul Desai, Apple’s vice president of Health and Fitness, called a dedicated research study “the most thoughtful way to understand how people engage with, comprehend, and evaluate health AI.”
The design choice worth noting is that the study measures trust and comprehension, not only accuracy. A correct answer that a person misreads, or an unsupported answer that a person trusts, are both failures. A study of this size, with ratings tied to specific answers, could give the field real evidence on which kinds of grounding people notice and act on.
/ 03The traceability angle
Citations are the visible part of grounding. They tell a reader that a source exists; they don’t by themselves show which sentence of the answer rests on which passage, or whether the answer changed because of the reader’s own data. Once personal health data such as lab results enters the prompt, an answer has two kinds of input: published evidence and the individual’s readings. For an answer to be checkable, both need to be recorded and linked to the claims they support.
That matters for the study itself. If a participant rates an answer as unhelpful or confusing, researchers will want to know what that answer was built from: which sources, which shared metrics, which guardrail fired. Answers that carry that structure can be audited and compared; answers that carry only a list of links are harder to learn from. The same holds when a patient brings an AI answer to a clinician, who needs to see quickly what it was based on.
/ 04What to watch
- Published findings. Apple frames the study as feeding the wider scientific conversation. Whether results on trust, comprehension and the effect of personal data are published, and in what detail, will determine its value to the rest of the field.
- Product follow-through. TechCrunch notes it is unclear whether Apple plans to bring AI health answers to the Health app or Siri.
- Data practices as a benchmark. No training on study data, anonymization before upload and 24-hour deletion of queries at OpenEvidence are concrete commitments that other consumer health AI products will be compared against.
A citation tells you a source exists. Traceability tells you which claim came from it, and which of your own numbers changed the answer.
Apple is right that the question is no longer whether people will ask AI about their health, but whether they can tell a grounded answer from a fluent one. A citation under each answer is the minimum. The harder part is showing which claim came from which page, and how a person's own data changed the answer.
In a knowledge graph, each statement in an answer is a node linked to the guideline or paper passage it derives from, and any personal reading it used (resting heart rate, a lab value) is linked as an input with its date. A clinician can then ask: which recommendation depended on this patient's labs, and which source supports it? That is how our cardiovascular build works, with 145 linked concepts and 284 recommended actions, each pointing back to the page it came from, running behind the firewall so patient data stays put.
How we’d model it
Health answer—cites→Journal articleAnswer claim—derivedFrom→Guideline passageAnswer claim—usesInput→HealthKit readingParticipant rating—evaluates→Health answer
- Apple launches research study to explore the use of AI for health — Apple Newsroom apple.com ↗
- Apple launches new app to study how people use AI for health — TechCrunch techcrunch.com ↗
See how Aktiver turns raw data into knowledge graphs your AI can reason over — deterministic, traceable, behind your firewall.