Connect with us

NEWS

Apple Puts OpenEvidence in a 200,000-Person Health Study

Apple’s new health study app runs on OpenEvidence, the clinical AI most U.S. doctors already use, and caps enrollment at 200,000 iPhone 15 Pro users.

Published

on

Apple opened the Apple Health AI Usage Study on November 15, 2026, with a dedicated iPhone app for everyday medical questions.

The answers come from OpenEvidence, a clinical AI that Apple says is widely used by healthcare professionals, and they arrive with linked peer-reviewed citations. Enrollment is capped at 200,000 U.S. adults, 21 and older, on an iPhone 15 Pro or later running iOS 26.

OpenEvidence Is the Engine Doctors Already Use

Apple did not build a new medical model for this trial. It plugged in OpenEvidence, a point-of-care search engine that is free for verified U.S. clinicians and locked behind an NPI check on the App Store.

The company’s own listing for the OpenEvidence NPI-gated clinician app says it is used daily by the majority of U.S. physicians, and that more American doctors use it than all other AI platforms combined. It claims licensed full text from The New England Journal of Medicine, JAMA and its 11 specialty journals, Nature, Cochrane, and the National Comprehensive Cancer Network, plus 35 million+ peer-reviewed papers and 300+ journals. It also lists 10,000+ hospitals and medical centers, including Mayo Clinic, Mount Sinai, Cedars-Sinai, Memorial Sloan Kettering, and Sutter Health.

That is a doctor product. Apple is now testing a consumer-facing version of the same cited stack, in an Apple-designed research shell, on phones that already hold HealthKit data.

THE TWO APPS SIDE BY SIDE

DetailOpenEvidence for cliniciansApple Health AI Study App
Who can use itVerified U.S. clinicians, NPI requiredU.S. adults 21+, iPhone 15 Pro or later, iOS 26
What grounds an answer35 million+ papers, 300+ journals, FDA and CDC sourcesOpenEvidence replies with linked peer-reviewed citations
Personal dataClinical detail the physician types inOptional HealthKit, including resting heart rate, activity, sleep, lab results
Reach named by the companyMajority of U.S. physicians daily; 10,000+ hospitalsUp to 200,000 people in one study
App Store score4.9 stars from 12K ratingsNew listing, not enough ratings to show a score
Stated limitHIPAA-compliant clinical support, not a consumer diagnosis appDoes not diagnose or treat; not a substitute for a clinician

OpenEvidence, in a July 20, 2026 release on the Stanford-Harvard NOHARM study, said that when 101 board-certified U.S. physicians could pick any tool, they reached for OpenEvidence in 22.3% of responses, more than ChatGPT, Claude, Gemini, and every other external AI combined at 19.8%. The company also said those benchmarking studies are not peer reviewed. The 22.3 and 19.8 figures are choice shares in that arm, not a full accounting of every resource a doctor used.

A Free Study App That Talks Like a Clinic Chatbot

Apple’s newsroom describes the Apple Health AI Usage Study as a virtual research project on trust, how clearly people understand answers, response quality, and everyday health actions. The App Store page for the free Apple Health AI Study App is blunter. It asks people to help Apple build better health AI products.

Both lines are Apple’s. One is the study frame. The other is the product pipeline written into the listing.

People in the study type everyday health questions and get simplified answers with links to the papers behind them. They can rate replies. Sharing HealthKit is optional, and Apple says it wants to see whether personal data makes those replies feel more relevant and easier to follow. The app can also try to help someone prepare for a visit with their own doctor. Guardrails are meant to catch some concerning questions and point the person toward a provider or other help.

WHAT THE STUDY APP SAYS IT CAN DO

  • Symptoms: Ask about symptoms or conditions and get an answer written in plain language, with citations.
  • Your Health data: Optionally fold in HealthKit metrics so the reply sits next to resting heart rate, activity, sleep, or lab results you already track.
  • Visit prep: Use the chat to line up questions before you see your own clinician.
  • Plans: Ask the app to draft a personalized exercise or meal plan from the data you choose to share.

That is not a diagnosis. Apple’s footnote on the newsroom post, and the App Store listing, both say the app does not diagnose or treat any condition and does not replace a medical professional.

Who Can Join the Apple Health AI Usage Study?

The gate is narrower than a normal App Store download. Apple wants an Apple Intelligence-enabled iPhone, specifically iPhone 15 Pro or later, on iOS 26 or later, and a U.S. adult who is at least 21. The cap is 200,000 people.

THE ENROLLMENT GATE

  • Place and age: United States only, 21 and older.
  • Phone: Apple Intelligence-enabled iPhone 15 Pro or later, running iOS 26 or later.
  • Headcount: Up to 200,000 eligible people.
  • Consent: Download the study app, read the study details, and choose to enroll; HealthKit and survey answers stay under the person’s control.

A standard iPhone 15 does not make the cut on Apple’s wording. The newsroom does not spell out a publication date for results, and the App Store listing still has too few ratings to show a score, which fits a study that opened on November 15, 2026.

Star ratings inside the app will tell Apple whether people found an answer clear, detailed, or relevant. They will not, on their own, prove the answer was clinically right. That gap is built into the design: the study is about engagement and trust, with citations as the grounding layer, not a new drug trial.

The Heart Study Wrote Apple’s Health Playbook

Apple has used large public studies before it ships health features, and it has kept the research and the product validation on separate tracks. The irregular rhythm notification on Apple Watch leaned on the Apple Heart Study. The ECG app leaned on a much smaller dedicated trial.

FROM THE HEART STUDY TO A WRIST ECG

  1. November 29, 2017: The Apple Heart Study opens in the United States as a virtual screening trial for irregular pulse and atrial fibrillation.
  2. December 6, 2018: Apple ships the ECG app and the irregular rhythm notification after FDA De Novo classification, citing a Heart Study of over 400,000 participants for the notification feature.
  3. December 6, 2018: Apple also reports a separate ECG trial of around 600 people, with 98.3% sensitivity for atrial fibrillation, 99.6% specificity for sinus rhythm, and 87.8% of recordings classifiable.
  4. December 6, 2018: In the Heart Study sub-study Apple submitted toward the notification clearance, 80% of people who got a watch alert while wearing an ECG patch showed atrial fibrillation on the patch, and 98% showed atrial fibrillation or other clinically relevant arrhythmias.
  5. November 15, 2026: Apple opens the Health AI Usage Study with OpenEvidence, again as a virtual app study, this time on conversational answers rather than a wrist sensor.

Sumbul Desai, M.D., who is now Apple’s vice president of Health and Fitness, was already speaking for these heart features in 2018. The through-line is the method. Run a big, opt-in study on Apple hardware, keep the data inside a research frame, then decide what, if anything, becomes a shipping feature.

The ECG app classified a 30-second tracing. The irregular rhythm feature looked for five irregular checks over at least 65 minutes. This new study is measuring something softer: whether people trust a cited answer, understand it, and do something after they read it.

ChatGPT Already Reads Apple Health at Far Larger Scale

OpenAI, on July 23, 2026, said more than 300 million people turn to ChatGPT each week with health-related questions, and it began rolling out Health in ChatGPT to U.S. users with an Apple Health connection. That figure is people, on OpenAI’s own page, not a query count.

ChatGPT can, with permission, pull in Apple Health plus supported medical records so a later chat can sit next to labs, sleep, and activity. OpenAI says information connected in Health is not used to train its foundation models or to target ads. Early testers still asked most health questions outside the dedicated Health space: more than 70% of those chats happened elsewhere in the product, which is why OpenAI stopped forcing people into a separate room.

Put next to Apple’s 200,000-person cap, that is not a close race. It is a research sample beside a habit that already exists on the same HealthKit stores. The difference Apple is testing is the engine. OpenEvidence answers from licensed journals and shows the paper. ChatGPT answers from general models that have been tuned on health evaluations, and it already speaks to Apple Health at consumer scale.

People will still ask about back pain at 3 a.m. The interesting question in this study is whether a citation changes what they believe, and whether they take that answer into a clinic the next morning.

Study Data Will Not Train Any AI Model

Apple says study data will not be used to train AI models. People keep control over HealthKit metrics and survey answers. The company says it has security measures around the study data it holds. The App Store privacy nutrition label, for this listing, names User ID as data linked to the person for analytics and app functionality, and performance diagnostics as data not linked to identity.

That is a research-app posture, not a full public accounting of every processor in the stack. Apple’s newsroom does not describe a 24-hour deletion window, and it does not name a third-party cloud in the announcement, so those details stay out of this account.

At Apple, we believe responsible innovation starts with scientific rigor and clinical grounding. Health is deeply personal, and conducting a dedicated research study is the most thoughtful way to understand how people engage with, comprehend, and evaluate health AI in everyday life.

Sumbul Desai, M.D., vice president of Health and Fitness, Apple newsroom, November 15, 2026

Desai also said insights from the study will feed a wider scientific conversation and, in her wording, help Apple keep building trusted tools. The listing’s “help Apple build better health AI products” line sits next to that sentence without much strain. The study is how Apple wants to watch people use a cited health chatbot before that chatbot is a Siri feature, a Health tab, or nothing at all.

Wachter’s Case for One Shared Clinical Tool

Robert M. Wachter, M.D., professor and chair of medicine at UCSF, is the named clinician in Apple’s announcement, and he described the split that this partnership tries to close.

Patients and clinicians are both turning to AI, but the tools they use are being built and tested in separate worlds. When patients and doctors draw on versions of the same trustworthy tools, patients arrive prepared, conversations improve, and next steps happen seamlessly. Connecting these tools, and rigorously studying what works, is how we’ll get there.

Robert M. Wachter, M.D., professor and chair of the Department of Medicine at UCSF, in Apple’s November 15, 2026 announcement

OpenEvidence remains a clinician product with an NPI gate. The study app is a consumer research client on top of that stack, limited to 200,000 U.S. adults on iPhone 15 Pro or later. If the trial works as Wachter frames it, a patient could walk into clinic having asked a cited engine the physician already knows. That is the stake for OpenEvidence, and it is the file Apple is building on how those answers land.

The Apple Health AI Study App is free for eligible U.S. iPhones. It does not diagnose or treat, and Apple’s own footnote tells people to take health concerns to a clinician.

Disclaimer: This article is news reporting and analysis of Apple’s Health AI Usage Study and the OpenEvidence collaboration. It is informational only and is not medical advice, a diagnosis, a treatment plan, or a recommendation to join or avoid the study. Readers who have a health concern, or who are deciding whether to share HealthKit data with a research app, should speak with a licensed physician or other qualified clinician and read Apple’s informed consent materials before acting. Enrollment rules, app behavior, and any figures cited here reflect the company and registry materials available on November 15, 2026, and those details can change.

Harry is the editor of NEWS ANALYSIS. He writes across the publication's ten desks, with most of his time going to the stories where a number, a filing or a study decides the argument. His working rule is simple: read the source document before writing about it, and tell the reader plainly which parts are established and which are somebody's claim. He is responsible for the standards set out on this site's Editorial Standards and Fact Checking pages, and for correcting the record openly when the publication gets something wrong.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending