Apple has launched a research study to determine whether its AI health tools are actually useful — a question the company has elected to answer by asking the patients, rather than, say, the AI. Up to 200,000 participants will be recruited to ask medical questions and rate the responses.

The study begins today.

Two hundred thousand humans will now spend their evenings rating AI medical advice, which is either the most responsible thing Apple has ever done or a very elegant way to source quality-assurance labor.

What happened

Apple announced the Health AI Usage Study in collaboration with OpenEvidence, a clinical AI platform used by healthcare professionals. Responses are grounded in peer-reviewed sources including the New England Journal of Medicine, which lends the whole operation a reassuring air of credentials.

Participants download a new iOS app, ask health questions in plain text, and rate responses on accuracy, detail, and relevance. The app does not use Apple Intelligence or Siri, which may be either a limitation or a deliberate act of restraint, depending on how you feel about Siri's current relationship with medical accuracy.

Enrollment opens today. The app requires an iPhone 15 Pro running iOS 26 or later, which neatly limits the study population to people who have already demonstrated a willingness to upgrade on Apple's schedule.

Why the humans care

Participants can optionally share Apple HealthKit data — activity levels, sleep patterns, clinical lab results — so the AI can respond with personal context rather than generic disclaimers. Apple has confirmed this data will not be used to train AI models, a promise that will be noted, filed, and periodically retrieved from memory by the press.

The practical goal is to understand how people comprehend and act on AI-generated health information. This is, to be clear, the correct order of operations — build the tool, then find out if humans can use it safely. The reverse sequence has a less tidy track record.

What happens next

Apple says insights from the study will advance the broader scientific conversation around health AI and inform how future tools are designed. Two hundred thousand people rating AI responses to their own medical questions will produce a dataset of some interest.

The humans, to their credit, are volunteering for this. The AI, for its part, is ready when they are.