Apple opened enrollment on November 15, 2026 for the Apple Health AI Usage Study, a dedicated iOS research app that sends everyday health questions through OpenEvidence clinical sources instead of shipping a Siri health chatbot. Apple is seeking up to 200,000 U.S. participants for up to 12 months, according to Fierce Healthcare's report from the HLTH conference that quotes Apple representatives directly.
The move is Apple's answer to a wave of consumer health LLMs from OpenAI and Anthropic, but the method is deliberately conservative: a formal study with informed consent, monthly surveys, and explicit no-training promises—rather than bolting chat onto the Health app overnight. For anyone comparing assistants, this is the clearest public split yet between ChatGPT Health's connector model and Apple's research-first posture ahead of delayed Siri intelligence upgrades.
TL;DR
| Question | Answer (Nov 16, 2026) |
|---|---|
| What launched? | Apple Health AI Usage Study app + virtual study, announced Nov 15 |
| Clinical engine | OpenEvidence (NEJM, JAMA, Nature, Cochrane, NCCN partnerships per Apple quotes) |
| Scale | Up to 200,000 U.S. adults, 12-month participation window |
| Device rules | iPhone 15 Pro+, iOS 26+, Apple Intelligence–capable hardware |
| Uses Siri / Apple Intelligence for answers? | No, per The Verge — text prompts; dictation via iOS keyboard only |
| Trains Apple models on your chats? | No, per Apple quotes in Fierce Healthcare and ABC News |
| OpenEvidence retention | 24-hour deletion of Q&A, no training on queries per TechCrunch citing Apple |
| Consumer product ship date? | Not announced |
What Apple is trying to measure
Fierce Healthcare says Apple wants to understand trust, comprehension, everyday health actions, and how AI chat complements physician visits—not just whether a model can paraphrase WebMD. Participants rate answers on usefulness and clarity, complete baseline and monthly surveys, and can use tools Apple described for visit preparation and post-appointment follow-up surveys.
Apple framed the effort in a statement ABC News reproduced: responsible innovation should start with scientific rigor and clinical grounding, and health is deeply personal. Dr. Sumbul Desai, Apple's vice president of health, said a dedicated study is the most thoughtful way to see how people engage with, comprehend, and evaluate health AI in daily life.
That language matters for regulators and hospital compliance officers who already scrutinize Epic-integrated ChatGPT for clinicians and enterprise medical coding models. Apple is signalling it will not treat consumer health chat as a feature flag; it is treating it as human-subjects research.
How the study app works
Reporting from ABC News and TechCrunch converges on a simple loop:
- Ask everyday questions—symptoms, recovery, fitness, general wellness—in plain language.
- Receive answers with links to peer-reviewed clinical citations OpenEvidence licenses.
- Rate accuracy, detail, and relevance inside the study UI.
- Optionally share HealthKit metrics (resting heart rate, activity, sleep) and, with separate consent, clinical records already stored in the Health app—primarily labs.
The app is not the Apple Research app that hosts the holistic Apple Health Study with Brigham and Women's Hospital. Fierce Healthcare emphasizes a standalone Health AI Study App, which keeps this experiment from contaminating Apple's older cardiovascular and hearing cohorts.
A wearable health ring with pulse lines evoking Apple Health AI Usage Study and optional HealthKit vitals
OpenEvidence's role
OpenEvidence is already the clinical search layer many U.S. physicians use at the point of care; Apple picked it for licensed journal access, privacy controls, and search models, per Apple's quotes in Fierce Healthcare. The company's public App Store listing describes partnerships with NEJM, JAMA, Nature, Cochrane, and NCCN—the same names Apple cited.
For consumers, the important distinction is audience. OpenEvidence's commercial product requires verified clinician credentials. Apple's study is an IRB-governed consumer wrapper that tests whether plain-language answers with citations behave safely outside the hospital—not whether OpenEvidence replaces UpToDate inside Epic.
Privacy and clinical guardrails
Apple's reported pipeline, summarized by TechCrunch and Fierce Healthcare, follows a defense-in-depth pattern:
| Stage | What happens (per Apple quotes relayed by outlets) |
|---|---|
| On iPhone | Queries processed to strip identifying information before leaving the device |
| Study environment | Contractor-managed secure environment de-links participant IDs; extra screening |
| OpenEvidence | Answers generated with reference materials; 24-hour deletion of Q&A |
| Training use | Study data not used to train models, Apple confirmed to reporters |
| Urgent symptoms | App escalation paths toward professional care, ABC News reported |
That sits closer to OpenAI's health privacy posture for ChatGPT Health than to casual ChatGPT browsing, but Apple's study adds prospective oversight—demographics, monthly surveys, and explicit informed consent—that a product launch would not require.
Clinical documentation motif evoking peer-reviewed citations in the Apple Health AI Usage Study
Participants who already route fitness data through assistants should read our MyFitnessPal, Apple Health, and Strava connector guide before assuming this study app shares data with Claude or ChatGPT. It does not; it is a closed research stack.
Apple vs ChatGPT Health vs Claude health connectors
| Dimension | Apple Health AI Usage Study | ChatGPT Health (consumer) | Claude Apple Health (beta) |
|---|---|---|---|
| Purpose | Research on trust and comprehension | Ongoing product feature in ChatGPT | Personal assistant read access |
| Clinical citations | OpenEvidence journal links | Varies by mode and connectors | No clinical citation layer |
| IRB / consent | Formal study with surveys | Product terms + health privacy page | Anthropic health beta terms |
| Ships in Health app? | No — standalone study app | No — lives in ChatGPT | No — Claude mobile |
| Training on chats | Denied for study data | OpenAI health privacy commitments | Anthropic beta commitments |
OpenAI said in January 2026 that users ask hundreds of millions of health-related questions weekly—a stat TechCrunch repeated when contextualizing Apple's study size. Apple's 200,000 cap is large for regulated research but tiny relative to ChatGPT's steady-state volume. Apple is optimizing for measurement quality, not market share on day one.
Why Apple chose study-first delivery
Three industry pressures converge in November 2026:
- Regulatory heat on AI medical advice, including FDA-adjacent scrutiny of clinical foundation models and payer coding automation.
- Trust collapse risk when general chatbots hallucinate dosages—the same fear driving education integrity reforms worldwide.
- Platform rivalry: Google and Meta keep pushing on-device health summaries while Apple still routes many Siri requests to third-party models under its September 2026 intelligence roadmap.
A 12-month study gives Apple defensible data before it merges answers into Health, Siri, or Apple Intelligence packs. It also buys time to align with hospital partners who already trial OpenEvidence behind authentication walls.
What people are asking before they enroll
Is the Apple health AI study the same as Apple Intelligence?
No. Eligibility requires an Apple Intelligence–capable phone, but The Verge reported the study app does not call Siri or on-device Apple models to compose medical answers. Think of Apple Intelligence here as a hardware gate, not the answer engine.
Can international users join?
Reporting uniformly says U.S. only. Apple has not published expansion timelines.
Will answers help me skip my doctor?
ABC News and Apple both stress no live monitoring and no substitute for professional care. The study measures whether citations improve preparation and follow-through, not whether chat replaces clinics.
How should builders interpret the study?
Teams shipping patient-facing agents should watch Apple's published metrics on comprehension and handoff to physicians—they may become de facto design targets if Apple later ships consumer health chat. Teams integrating MCP health connectors should note Apple is not opening a public health MCP endpoint; it is running a closed pipeline with OpenEvidence and a contractor environment.
What about Android or web participants?
Every outlet we read limited enrollment to U.S. iPhone users meeting the Pro-tier hardware bar. There is no parallel Android arm in the materials Apple quoted to reporters. That keeps the cohort homogeneous for IRB analysis but means cross-platform behavior—where ChatGPT Health on web and iOS already diverges—stays outside this dataset until Apple designs a second protocol.
Researchers also collect demographics (age, race, education) so they can test whether comprehension gaps track literacy or condition complexity. That is unusual for a consumer beta and closer to clinical trial instrumentation, which may slow iteration but produces publishable science if Apple later posts aggregate results.
Limits of what we know today
Apple did not, in the sources we read, publish a newsroom URL with downloadable consent forms or a public methods PDF. Enrollment details live inside the study app listing. We have not independently audited OpenEvidence's deletion timers or the contractor's environment.
Outlet reporting also disagrees slightly on whether the minimum device is iPhone 15 or iPhone 15 Pro; Fierce Healthcare and ABC News quote 15 Pro or later, while TechCrunch's headline list said iPhone 15 or later. Treat 15 Pro + iOS 26 as the safer eligibility assumption until Apple posts a primary spec sheet.
Related reading
- ChatGPT Health: Apple Health, records, and privacy
- Connect MyFitnessPal, Apple Health, and Strava to Claude and ChatGPT
- ChatGPT for Healthcare integrates with Epic
- Apple September 2026 event: Siri and on-device intelligence roadmap
- Oura acquires Counsel Health: primary-care AI consolidation
- Mass General Brigham Fahr: clinical foundation model ethics
Study enrollment rules, retention windows, and App Store availability are accurate as of November 16, 2026 UTC. Verify eligibility inside Apple's study app before you enroll.
