Apple and OpenEvidence study AI's role in personal health

November 18, 2026
Apple and OpenEvidence study AI's role in personal health
AI
News

Apple and clinical AI company OpenEvidence have launched a large-scale study involving up to 200,000 people in the United States to investigate how consumers use artificial intelligence to understand their health. The research will examine whether medically grounded AI can make complex information more accessible, help patients prepare for consultations and support better-informed health decisions. It also addresses a question that healthcare providers increasingly face: what happens when patients begin relying on conversational AI before speaking to a clinician?

Patients are increasingly using generative artificial intelligence to interpret symptoms, understand medical terminology and make sense of laboratory results. For healthcare professionals, this creates a new reality in which patients may arrive at appointments with detailed explanations or recommendations generated by systems operating outside the healthcare organisation. Whether these interactions improve understanding or introduce additional uncertainty remains insufficiently understood.

Apple is now attempting to investigate that question at scale. On 15 November 2026, the company announced the Apple Health AI Usage Study, developed in collaboration with OpenEvidence, a clinical AI platform used by healthcare professionals. The virtual study will enrol up to 200,000 eligible adults across the United States and examine how people assess AI-generated medical information, what makes them trust the responses they receive and whether those interactions influence everyday health-related behaviour.

The initiative represents a notable development in consumer health technology. Apple has spent years building an ecosystem around health measurements, wearable devices and research applications. OpenEvidence, meanwhile, has developed a conversational platform that provides medical information supported by scientific literature. Their collaboration brings these two areas together in an attempt to understand how AI might help consumers interpret health information that would otherwise require searching through multiple sources or consulting a healthcare professional.

Unlike conventional assessments of AI systems, which often focus on the accuracy of answers to medical questions, this study places considerable emphasis on the person receiving the information. An answer can be scientifically accurate yet still be misunderstood, presented without sufficient context or interpreted as advice that is more definitive than the evidence supports. Apple wants to understand these interactions before drawing conclusions about how conversational AI should be incorporated into consumer health services.

Connecting medical evidence with personal health data

Participants will use a dedicated iPhone application developed for the study. They can ask questions about their health, explore explanations of medical subjects and review references to peer-reviewed research supporting the responses. The underlying AI technology comes from OpenEvidence, which draws on established medical sources including The New England Journal of Medicine, JAMA and Nature.

The application also allows participants to evaluate the responses they receive. Researchers will examine factors including perceived quality, relevance, comprehension and trust. According to reporting by Fierce Healthcare, participants will receive baseline and monthly surveys, enabling researchers to investigate how people use the information and what actions they subsequently take.

One of the most significant elements is the optional integration of personal health information. Participants may choose to share selected data through Apple's HealthKit framework, including resting heart rate, physical activity, sleep measurements and laboratory results. Researchers will examine whether access to these measurements makes AI-generated explanations more relevant and easier to understand.

This distinction is important. Explaining what an elevated resting heart rate might mean is different from interpreting a person's heart rate alongside their activity patterns, sleep history and other available measurements. Personal information may provide useful context, but it also introduces additional complexity. Measurements collected through consumer devices are not necessarily sufficient for clinical interpretation, and AI systems may lack important information about medical history, medication or underlying conditions.

The study therefore does not assume that personalisation automatically produces better advice. Instead, it aims to investigate whether additional information improves the usefulness and comprehension of responses.

Another area of research concerns preparation for medical appointments. Participants will be able to use the application to organise health questions and prepare for conversations with their own doctors. Follow-up surveys are intended to explore whether these interactions influence how prepared patients feel and how they experience consultations.

Robert M. Wachter, professor and chair of the Department of Medicine at the University of California, San Francisco, highlighted the growing separation between AI systems designed for patients and those developed for clinicians. Both groups are increasingly using the technology, but their tools are often evaluated independently. Better alignment between them could potentially improve communication, although this remains to be demonstrated through research.

A new challenge for healthcare providers

The research comes as conversational AI is becoming a common source of health information. According to Healthcare Dive's reporting on the announcement, the initiative was presented in Las Vegas, where Apple executives emphasised the need to understand consumer trust, response quality and the potential influence of AI on behaviour.

For healthcare organisations, these developments extend beyond the introduction of another digital health application. Patients can already consult AI systems independently, often before contacting their healthcare provider. They may use the technology to understand a diagnosis, evaluate treatment information or formulate questions about symptoms. This creates opportunities for more informed conversations, but also raises concerns about inaccurate information, inappropriate reassurance and the possibility that users may place too much confidence in automated responses.

The difference between information and medical advice is particularly relevant. A conversational system may explain established medical knowledge without being able to assess an individual's complete clinical circumstances. Even when responses include references to scientific publications, the interpretation of those sources may be incomplete or inappropriate for a particular patient.

Apple acknowledges these limitations. The research application includes safeguards intended to identify certain concerning questions and recommend that participants contact healthcare professionals or other appropriate services. The company explicitly states that the application is not intended to diagnose or treat medical conditions and should not replace professional medical advice.

Privacy and data governance are also central to the research design. Apple says study information will not be used to train AI models, and participants can decide whether to share selected HealthKit measurements and survey responses. According to the company's explanation of its privacy arrangements, the study includes security measures intended to protect participant information.

Further details reported by TechCrunch indicate that identifying information is removed from questions before they are processed through the research infrastructure, while OpenEvidence deletes question-and-response data after 24 hours.

These measures are particularly relevant because the combination of conversational AI and personal health measurements introduces sensitive information into interactions that may occur outside traditional healthcare systems.

The study also has limitations that will matter when interpreting its eventual findings. Participation is restricted to US residents aged 21 and older who have an Apple Intelligence-compatible iPhone running iOS 26 or later. Participants must voluntarily enrol, meaning the study population may differ from people with limited digital access, lower levels of health literacy or less experience using technology.

The research is designed to investigate trust, comprehension and health-related behaviour. It should not be interpreted as a clinical trial demonstrating that conversational AI improves diagnosis, treatment effectiveness or patient outcomes.

From consumer technology to clinical communication

The broader significance of the Apple and OpenEvidence collaboration lies in the changing relationship between patients, medical information and healthcare professionals.

Digital health services have traditionally focused on collecting information, enabling remote monitoring or providing access to medical records. Conversational AI introduces an additional layer: technology that attempts to explain information and respond to questions in natural language. This could make complex medical knowledge more accessible, particularly for patients trying to understand unfamiliar terminology or prepare for appointments.

Whether this translates into better healthcare depends on more than the quality of the underlying language model. Patients must be able to recognise uncertainty, distinguish general explanations from personalised medical advice and understand when professional assessment is necessary. Healthcare professionals, in turn, may need to adapt to consultations in which AI-generated information forms part of the discussion.

For hospitals and health systems, this raises practical questions about patient education, digital literacy and the integration of consumer-generated information into existing care pathways. It also creates a potential role for clinically grounded AI applications that support patients between appointments without attempting to replace the expertise of healthcare professionals.

Although the Apple study is being conducted exclusively in the United States, its findings could be relevant internationally. European healthcare organisations are working to improve access to health information while developing governance frameworks for AI and the secondary use of health data. Understanding how patients interpret AI-generated explanations may become increasingly important as these systems enter everyday healthcare.

The study has only recently begun, and no findings about effectiveness have been published. Its contribution will ultimately depend on whether the researchers can demonstrate not simply that people appreciate conversational AI, but that they understand the information it provides and use it appropriately.

For healthcare providers, that distinction is fundamental. The ability to generate convincing medical explanations is already widely available. Establishing whether those explanations help patients make better-informed decisions is a different and considerably more demanding task.

From innovation to healthcare transformation

These developments will also be part of the broader international conversation at the annual ICT&health World Conference 2027, taking place from 26–28 January in Maastricht, the Netherlands. Bringing together healthcare leaders, clinicians, researchers, policymakers, governments and technology innovators from more than 50 countries, the conference addresses the challenges shaping the future of healthcare. From artificial intelligence, digital health and medical technology to interoperability, data sharing, workforce transformation, leadership and new models of care, the focus extends beyond technological possibilities to the organisational, financial and behavioural changes needed to make innovation work in everyday practice. Across more than 150 sessions featuring over 350 speakers, participants will exchange knowledge, share practical experiences and explore how healthcare systems can move from isolated initiatives to sustainable, large-scale transformation.

Add ICT&health on Google

Show more content from ICT&health in Google Search.

Add on Google

This topic will also have a prominent place at the ICT&health World Conference 2027. Want to be there and stay ahead of what’s next in healthcare? Reserve your ticket today.