Apple and OpenEvidence Launch AI Health Chat, Seeking Up to 200,000 Participants
Apple has launched an app with an AI health chat and is recruiting up to 200,000 volunteers in the US for more than 12 months. OpenEvidence builds the answers to questions like "why am I sleeping badly" from articles in the New England Journal of Medicine, JAMA, and Nature. Participants rate the answers and complete surveys, which lets Apple test whether people understand and trust AI medical advice. For product teams, the takeaway is to measure trust before scaling AI answers in sensitive areas.
AI-processed from TechCrunch; edited by Hamidun News
Apple has launched an app that lets people ask an AI about health, and it wants to recruit up to 200,000 volunteers in the US over more than 12 months. The answers are not generated by Apple itself but by a partner, the medical AI company OpenEvidence, which draws on articles from the New England Journal of Medicine, the Journal of the American Medical Association (JAMA), and Nature, all leading clinical journals. This is not a finished AI doctor but a large experiment: Apple wants to understand how people use such answers, whether they understand them, and whether they trust them.
Why Apple needs the study
Millions of people already ask chatbots about symptoms, test results, and medications. OpenAI alone reported more than 300 million health-related queries per week this year. OpenAI and Anthropic have already released chatbots with dedicated health features. Apple is taking a different route: it is first collecting data on how people behave and only then deciding what to build into its products.
The app looks familiar. A user types a question, as in ChatGPT or Claude, and gets an answer. The difference is in the sources: the answers are grounded in clinical data. Apple says the app will simplify answers to everyday medical questions so that an ordinary person can understand them.
Participants can rate the answers, leave feedback, and complete a survey once a month. This is the core of the project: Apple is studying not only the questions themselves but also how people perceive the answers.
Users can optionally connect HealthKit data: resting heart rate, activity, sleep, and lab results. Doing so makes the advice more precise. A question about fatigue, for example, is addressed with consideration of how the person has been sleeping over recent weeks. The app also helps people prepare for a doctor's visit and directs them to specific specialists when a question involves worrying symptoms.
Conditions and privacy
Participants must be US residents over 21. They need an iPhone 15 or newer, iOS 26 or later, and Apple Intelligence turned on. The app can be downloaded now, but only in the US for the time being.
The most sensitive issue here is data. Apple states that information is de-identified before being sent to a secure third-party cloud provider and is not used to train any AI models. According to the company, OpenEvidence deletes queries and answers after 24 hours. These promises are hard to verify from the outside, so for now they remain the company's claims. For medical data they matter a great deal: because of privacy concerns, many people do not trust AI with health questions.
Apple's Vice President of Health and Fitness, Sumbul Desai, explains the approach this way: health is a very personal subject, and a dedicated study conducted with scientific rigor is the most thoughtful way to understand how people use medical AI, how they understand it, and how they evaluate it. The results, she said, should feed into the broader scientific conversation.
What this means for developers and product managers
First, Apple has chosen a strategy different from OpenAI's and Anthropic's: a bet on verifiable sources and on measuring trust. For those building medical or health-adjacent products, the signal is that journal references and clinical data become part of the product, not decoration.
Second, the choice of partner is telling. Apple did not try to build everything on its own model right away. Instead it took a specialized service with an existing literature base. A major platform is testing the idea on someone else's technology before building it into its operating system.
Third, the story has honest limits. It is unclear whether such answers will appear in the Health app or in Siri: Apple has not promised that. The study is limited to the US, and only owners of newer iPhones with Apple Intelligence enabled can take part, so the sample will not be fully representative. There is no data yet on the accuracy of the app's answers; what is known is only which sources it draws on and how the process is structured.
Still, the potential is large. If in a year Apple has a substantial dataset on which questions people ask, where they need simplified explanations, and when they decide to see a doctor, it will strongly shape the design of future AI features in Health.
The practical takeaway: if you are building an AI assistant for a sensitive domain, build in user-driven evaluation of answers, short data retention periods, and clear routes to a live specialist.
- Nov 21, 2026OpenAI GPT-6 Cyber Found About 35 Flaws in iPhone, Mac and Other Apple Devices
- Nov 21, 2026Apple's macOS 27.3 beta 2 will let you limit AI agents' access to Mac files
- Nov 21, 2026Apple Is Building an AI Agent for Mac While Making It Harder for Outsiders to Get In: 14 Clicks Instead of 4
- Nov 21, 2026Apple Is Recruiting 200,000 Volunteers to Test a Health AI Chatbot
Key Questions Answered
Why Apple needs the study?
Millions of people already ask chatbots about symptoms, test results, and medications. OpenAI alone reported more than 300 million health-related queries per week this year. OpenAI and Anthropic have already released chatbots with dedicated health features. Apple is taking a different route: it is first collecting data on how people behave and only then deciding what to build into its products. The app looks familiar. A user types a question, as in ChatGPT or Claude, and gets an answer.
What are the key details on Conditions and privacy?
Participants must be US residents over 21. They need an iPhone 15 or newer, iOS 26 or later, and Apple Intelligence turned on. The app can be downloaded now, but only in the US for the time being. The most sensitive issue here is data. Apple states that information is de-identified before being sent to a secure third-party cloud provider and is not used to train any AI models. According to the company, OpenEvidence deletes queries and answers after 24 hours.
What this means for developers and product managers?
First, Apple has chosen a strategy different from OpenAI's and Anthropic's: a bet on verifiable sources and on measuring trust. For those building medical or health-adjacent products, the signal is that journal references and clinical data become part of the product, not decoration. Second, the choice of partner is telling. Apple did not try to build everything on its own model right away. Instead it took a specialized service with an existing literature base.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.