AI

Apple wants 200,000 iPhone owners to grade AI answers about their health

Adrian Kessler
Add us on Google

Apple has put a health chatbot on the iPhone, but it sits inside a research study rather than in Siri or the Health app. The new Apple Health AI Usage Study app lets participants type everyday medical questions, get answers in plain language with links to peer-reviewed clinical sources, and then grade those answers for accuracy, detail and relevance. Apple is recruiting a very large panel of volunteers to do it.

The point is to find out whether people understand and trust what a health AI tells them, and whether it changes what they do next. That question already matters to anyone who asks a chatbot about a rash or a lab result: one in three American adults use AI chatbots for health information, as CNET noted, and OpenAI has said its users send it hundreds of millions of health-related questions every week.

Apple wants up to 200,000 participants, and the answers they get do not come from Apple’s own model. They are generated by OpenEvidence, a clinical AI platform widely used by doctors, which draws on sources such as The New England Journal of Medicine, the Journal of the American Medical Association, Nature and the Cochrane database of systematic reviews. That is a far narrower diet than a general chatbot trained on the open web, and every answer shows its citations.

Participants can also choose to share HealthKit data such as resting heart rate, sleep, activity or lab results, so a question like “how can I improve my heart health” gets an answer built around their own numbers. The app can also help people prepare for a visit to their own doctor, and it points them to a clinician when a question sounds like it needs one.

Apple says study data will not be used to train AI models. According to CNET, each question is first scrubbed of identifying details on the iPhone, then passes through a cloud environment run by a third-party contractor, UST Global, where it is separated from the participant’s ID before it reaches OpenEvidence. OpenEvidence deletes questions and answers within 24 hours. Apple’s on-device models only sort questions into categories, and the app is not connected to Siri.

The limits are real. The study is open only to adults aged 21 and over who live in the United States, with no word on other countries, and it needs an Apple Intelligence iPhone, meaning an iPhone 15 Pro or later on iOS 26, which leaves out a large share of the iPhones people actually use. It is a research app, not a medical device, and Apple says it does not diagnose or treat anything. Its safety guardrails flag worrying questions and suggest a doctor or other resources, but nobody watches conversations live, and study staff will not contact a participant who asks about self-harm, CNET reported. Participants are also asked for their race, age and education level, which Apple says will show how communication needs differ between groups.

“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,” said Sumbul Desai, Apple’s vice president of Health and Fitness. Robert Wachter, chair of the Department of Medicine at UCSF, made the case in Apple’s announcement: patients and clinicians are both turning to AI, he said, “but the tools they use are being built and tested in separate worlds.”

Apple announced the study on November 15, and the app is available now on the US App Store. Each participant can stay in for up to 12 months, with a baseline survey, monthly check-ins and extra surveys around doctor’s visits, and people who stop taking part entirely may be removed. Apple has set no date for publishing the results and has not said whether any of this will reach the Health app or Siri.

For now, the only way to ask Apple’s health app a medical question is to sign up as a research subject.

Tags: , , , , ,

Add us on Google

Discussion

There are 0 comments.