Apple Launches Study App to Test How People Use AI for Health Questions

A research study, not a product launch
Apple's new app is built to learn how people use AI for medical questions, not to sell a feature.
Mark

Why does it matter that Apple is calling this a study and not a feature?

Mimi

Because a study has defined participants, surveys and feedback built in. Apple wants to learn how people judge these answers before deciding what to build.

Mark

What separates it from asking ChatGPT a health question?

Mimi

The answers are said to rest on clinical sources like NEJM, JAMA and Nature, and users can optionally link their Apple Health data for personalization.

Luke

Careful with "rest on." That's Apple's description. The reporting doesn't show how the app selects or cites those sources, or how often it gets things right.

Mark

And the privacy claims?

Mimi

Data won't train AI models, it's anonymized before going to a cloud provider, and OpenEvidence deletes questions and responses after 24 hours.

Luke

Those are stated commitments. Nothing here says they've been independently verified, and anonymization is a claim, not a proof.

Mark

The scale seems large.

Mimi

Up to 200,000 people over more than 12 months, US only, 21 and older, on iPhone 15 or newer.

Luke

"Up to" matters. That's a ceiling, not an enrollment figure. And the 300 million weekly questions number is OpenAI's own statement.

Mark

Will this reach Siri or the Health app?

Mimi

Nobody has said. That's still open.

  • Apple is quietly testing AI health answers on up to 200,000 iPhone users, pressing into a space where rivals have already moved fast and at enormous scale.
  • The study asks people to type real medical questions and optionally link sleep, heart rate and lab data, raising the stakes for what the machine might get wrong.
  • Strict privacy promises are being put forward as the foundation: no AI training on personal data, anonymization before cloud handling, and deletion of chats within 24 hours.
  • The path forward is still open, with a year-long run and monthly surveys ahead, and no clear word yet on whether the feature will reach the main Health app or Siri.

In a moment when millions of people are already turning to chatbots with their worries about bodies and symptoms, Apple has chosen to slow the conversation down with a research study rather than a product release. Working with the healthcare AI firm OpenEvidence, it will test whether answers anchored in established medical literature can genuinely serve ordinary people in the United States. The question at its heart is an old one in new clothing: how much should we trust guidance that arrives instantly, and what does it take to earn that trust over time.

Apple has opened a new research app to iPhone users in the United States, and its aim is to learn rather than to sell. The Apple Health AI Usage Study is designed to see how people use artificial intelligence when they have a health question, arriving at a time when OpenAI has reported more than 300 million health-related questions to its chatbots every week. Apple is taking a more deliberate path, partnering with OpenEvidence to test whether answers grounded in real clinical sources can truly help.

The study will enroll up to 200,000 adults aged 21 and older over more than a year. Participants type questions the way they would into any chatbot, and the answers draw on journals such as the New England Journal of Medicine, JAMA and Nature. Those who choose can link their Apple Health information, including sleep, heart rate, activity and lab results, so responses feel personal. The app can also help people prepare for a doctor's visit or point them toward specialists when something seems serious. Participants rate answers and complete monthly surveys.

Apple places its firmest commitments on privacy. Personal data will not train any AI model, information is anonymized before reaching a third-party cloud provider, and OpenEvidence deletes questions and responses after 24 hours. Dr. Sumbul Desai, Apple's vice president of Health and Fitness, described the effort as a matter of responsible method, grounded in science. What comes next remains unsettled, including whether this feature will eventually appear in the Health app or Siri, but the findings will reveal how far clinically grounded AI guidance can safely grow.

Apple has put a new app in front of iPhone users in the United States, and its purpose is research rather than a product launch. It is called the Apple Health AI Usage Study, and it is built to find out how people use artificial intelligence when they have a medical question.

The timing follows a number from OpenAI, which recently said users put more than 300 million health-related questions to its chatbots every week. Anthropic and OpenAI both already run chatbots that field health queries. Apple says it is taking a different route. It has partnered with OpenEvidence, a healthcare AI company, to test whether answers tied to real clinical data can actually help people.

The study is limited in who can join. It is open only in the US, and Apple plans to enroll up to 200,000 people for more than 12 months. Participants must be 21 or older and own an iPhone 15 or newer. The phone must run iOS 26 or later with Apple Intelligence switched on.

Inside the app, a user types a health question much as they would in ChatGPT or Claude. The app tries to return a plain, easy-to-follow answer to an everyday medical question. Those answers draw on clinical sources, among them The New England Journal of Medicine, JAMA and Nature.

Users can also link the app to their Apple Health data through HealthKit. That would let it draw on sleep patterns, resting heart rate, daily activity and lab results to personalize what it says. Apple says the app can help someone prepare for a doctor's visit, and can point to specific providers when a medical concern looks serious. Participants can rate answers, leave feedback, and will be asked to complete monthly surveys.

Privacy is where Apple makes its firmest commitments. Personal data collected through the app will not be used to train any AI models. Before information goes to a secure third-party cloud provider, it is anonymized so it cannot be tied back to the user. OpenEvidence deletes users' questions and the AI's responses after 24 hours.

Dr. Sumbul Desai, Apple's vice president of Health and Fitness, framed the study as a matter of method. She said the company believes responsible innovation begins with science and clinical grounding. Health, she said, is deeply personal, and a dedicated research study is the most thoughtful way to see how people engage with and evaluate health AI in daily life. Apple says what it learns will help it build trusted tools that support people on their health journey.

What the study will become is not settled. It is unclear whether Apple intends to bring this answering feature to the main Health app or to Siri. For now, the 12-month run and its findings will show how far AI health guidance grounded in clinical literature can be trusted to scale.

Responsible innovation starts with science and clinical grounding, and a dedicated study is the most thoughtful way to see how people engage with health AI.
— Dr. Sumbul Desai, Apple VP of Health and Fitness (paraphrased)
Quer a matéria completa? Leia o original em India Today ↗
Fale Conosco FAQ