This post is written in English.
Apple has launched a study to find out how people use conversational AI for health questions, and how useful the answers are. In its announcement, as reported by The Verge, the company calls it the Apple Health AI Usage Study. It runs with OpenEvidence, a clinical AI platform that healthcare professionals use. OpenEvidence generates answers from medical sources such as The New England Journal of Medicine.
How the study works
Participants install a new iOS app, ask health questions, and rate the responses on accuracy, detail, and relevance. The app gives easy-to-read answers with links to peer-reviewed clinical sources.
Enrollment opened on the day of the announcement, and Apple is looking for up to 200,000 people. To take part, you need to be 21 or older and based in the US. You also need an iPhone 15 Pro running iOS 26 or later.
The app is text-based. You can dictate using the iOS keyboard, but the app isn't integrated with Siri or Apple Intelligence, and Apple says it doesn't use Apple Intelligence.
The health data question
During sign-up, participants can choose to share Apple HealthKit data such as activity, sleep, or clinical lab results. Apple says this would give responses more context about the person's own health.
Apple also says it has taken steps to keep that information confidential, and that the data collected won't be used to train AI models. Those are Apple's claims. The announcement doesn't describe how those steps work in practice, so readers will have to take the company at its word for now.
What Apple says it wants
Apple says the goal is to improve how AI health tools are designed, so they give information that is easier to understand and act on. Sumbul Desai, MD, Apple's vice president of Health and Fitness, said the study is "the most thoughtful way to understand how people engage with, comprehend, and evaluate health AI in everyday life."
What's not clear yet
The announcement doesn't say how long the study will run, how the ratings will be published, or whether the results will be shared outside Apple. It also doesn't say which model generates the app's answers, or how OpenEvidence's role differs from Apple's.
My read
Asking users to rate answers is a sensible way to measure whether health AI is accurate and clear. A rating only tells you how people judged a response, though, not whether the response was correct. Anyone interested in the results should look for the methods and findings, not just the sample size. The data-sharing terms also deserve attention. Sharing lab results is a bigger step than rating a chatbot's answer, so the privacy promises need to be checked against the study's published details once they exist.
← All posts


