Navigating AI in Healthcare as a CitizeN

Navigating AI in Healthcare as a Citizen

By Dr. Sanaz Jamalzadeh

For years, the conversation around artificial intelligence (AI) in healthcare has focused on institutions. We have heard about AI for hospitals, clinicians, diagnostics, and health system efficiency. These are important developments, and they deserve attention, but it is no longer the full picture. 

A quieter and perhaps more profound transformation is now taking place at the level of everyday life. More and more people are generating health-related data continuously. It comes from smartwatches and fitness trackers, patient portals, laboratory reports, electronic health records and increasingly, digital conversations themselves. At the same time, AI systems are beginning to interpret these streams, summarize them, and turn them into recommendations that feel immediate, personal, and actionable

This raises a question that deserves far more attention than it currently receives. How should ordinary people prepare to navigate an AI-enabled, data-rich health world?

Traditionally, healthcare interpretation happened primarily within clinical settings, where physicians and other professionals placed a lab result, symptom pattern, or imaging finding into a broader medical context. Today, that interpretive layer is increasingly moving closer to the individual. A person may first encounter meaning through a dashboard, an app, or a conversational AI tool before they ever speak to a clinician. In that sense, the next frontier of AI in healthcare is not only smarter systems for healthcare and medicine, but a new relationship between citizens and their own health data (1,2). Meaningful digital health depends not only on technology being available, but on people being able to access, understand, and use their own health information (3). 

The real challenge is judgment

Today the promise of AI is becoming inseparable from its risks. More interpretation does not automatically mean more understanding. AI can detect patterns, summarize trends, and generate plausible explanations, but most people are not clinicians, nor should they need to become clinicians in order to use digital health tools responsibly. The main question, then, is under what conditions AI helps people without confusing, misleading, or overwhelming them. Recent evidence suggests this concern is well founded (4,5). Generative AI often fell short of evidence-based standards when people used it to seek health screening information, even though some prompting strategies improved performance. Moreover, nonexperts may overtrust AI-generated medical advice despite low accuracy, especially when the answers are presented in confident and polished language. This combination is particularly important in healthcare, where persuasive wording can easily be mistaken for reliability. 

What people therefore need is not simply more data, instead they would need better judgment. A wearable trend, a risk score, or an AI-generated explanation is not the same as a diagnosis. In a data-rich world, one of the most important public health skills will be learning how to use AI as an aid to reflection and question-formulation, without confusing it for medical authority.

A new form of health literacy

This is why AI in healthcare must also be understood as a literacy challenge. Traditional health literacy meant being able to access, understand, and use health information. In the age of Generative AI, people also need to know how to interpret machine-generated health explanations. They need to ask what data a system is using, whether its output is grounded in validated medical evidence, whether uncertainty is being communicated honestly, and whether the tool is designed for education, coaching, or something closer to clinical decision assistance. 

Importantly, this cannot become the sole burden of the user. We should not build systems that require anxious, unwell, or time-pressured people to become experts in prompting or model evaluation just to stay safe. If patient-facing AI is to be genuinely useful, safeguards must be designed into the experience itself. Systems should make their limits visible, distinguish information from diagnosis, communicate uncertainty clearly, and know when escalation to a healthcare professional is necessary. Patient trust in medical AI is often shaped by factors such as performance, clinician involvement, and formal oversight or approval mechanisms (6,7). Patients also have a legal basis to ask for explanations of AI-driven medical recommendations. But having the right to an explanation is not the same as receiving one that is actually useful. Technical opacity, time pressures in clinical settings, and wide variation in health literacy all stand between patients and meaningful understanding. 

Addressing these challenges need action from multiple actors. Developers could design explanation systems with patient input from the outset, testing comprehension with actual patients rather than demonstrating compliance with legal standards alone. Health care institutions can also bridge this gap. Allocating time for AI discussions, training staff to support patients in navigating AI-driven recommendations, and establishing clear protocols could shift explanation from a compliance exercise toward genuine patient understanding (8,10). Policy makers could also support this by developing standards focused on comprehension and investing in digital health literacy programs (10).

What can citizens do next? 

Citizens do not need to become AI experts, but they do need a more active stance toward AI-generated health information. That starts with access. If people are expected to live in an AI-enabled health system, they should not only receive AI-supported outputs but also be able to access, understand, and use their own health data. This includes electronic health records, laboratory reports, imaging results, medication histories, and other parts of the medical record that increasingly shape diagnosis, treatment planning, and care navigation. Citizens should not have to remain passive recipients of conclusions made around them, but should be able to bring their own data into conversations and decisions about their care. 

Public healthcare systems and healthcare institutions therefore have an important responsibility to make patient data available in ways that are timely, understandable, and usable - without compromising privacy or security - so that access to data supports shared decision-making rather than remaining a purely technical or administrative function. Patient access to health data and the importance of shared decision-making play a major role in people-centred digital health.

Once citizens have access to their own health data, the next question is how to use AI well. The most constructive use of these tools is not to hand over judgment to them, but to use them for reflection, preparation, and question-formulation. People may use AI to summarize records, clarify unfamiliar terms, or prepare questions for a clinical visit, while still checking whether the output is evidence-based and whether uncertainty is clearly communicated. The goal is not blind trust, but a more deliberate form of use in which AI supports health understanding without displacing clinical care (4). 

A further and equally important possibility is that AI may help people make sense of their own data in context. Citizens may increasingly benefit from systems that show for instance how sleep, activity, symptom patterns, or clinical indicators compare with validated population benchmarks rather than presenting isolated numbers alone. That promise is real, but it depends on trustworthy infrastructure. Person-generated health data can support more proactive and personalized care only when people can retrieve and use their own records, and when those records can be securely exchanged and interpreted within governed systems rather than opaque consumer environments that are driven by tech companies. The underlying datasets must be representative, longitudinal, and carefully governed. Otherwise, apparently personalized or population-based insights may still rest on biased or incomplete comparisons. For that reason, the more promising future is not unrestricted “plug-and-play” health AI, but citizen-facing tools connected to trustworthy data infrastructures that combine personal relevance with transparent benchmarks and appropriate oversight (9).

Designing AI that patients can actually use

If patient-facing AI is to be genuinely useful, implementation must go beyond simply making these tools available or technically explainable. It starts with debate between clinicians, hospitals, public health organizations, and citizens on data ownership. Data is the bedrock that enables this technology. Citizen ownership of their own records, the possibilities with data collectives, and other mechanisms of opening up and sharing data and information between all kinds of institutions - with advancement in technologies and protocols that would make this possible is an important conversation and step towards an informed and healthier population. 

What matters is whether people can actually use AI outputs from their own data to make informed decisions about their care. The goal is that explanations should not focus only on how an algorithm works internally, but on what the system is recommending, how confident it is, what that confidence means in practical terms, and what alternative options or next steps remain available. Just as standards exist for medical devices, similar standards are needed for AI systems that communicate health information, with a clear chain of responsibility. 

Developers should design explanation systems with patient input from the outset, healthcare institutions should allocate time and support for AI-related conversations, and policymakers should promote standards centered on comprehension and digital health literacy rather than disclosure alone. In this way, trustworthy AI in healthcare becomes a shared responsibility across design, care delivery, and governance (10). While we have discussed citizen-level benefits of data-ownership and AI layer, the opportunities these same developments create for health institutions, clinicians, and public health are also significant. 

The next phase of digital health will then be defined by whether ordinary people access and use their own health data as part of care decisions and navigate AI-mediated health information with confidence, support, and meaningful understanding. If we get that right, AI can become a meaningful layer of assistance between people and the complexity of modern healthcare. 

References

1. Merrill, M. A., Paruchuri, A., Rezaei, N., Kovacs, G., Perez, J., Liu, Y., Schenck, E., Hammerquist, N., Sunshine, J., Tailor, S., Ayush, K., Su, H.-W., He, Q., McLean, C. Y., Malhotra, M., Patel, S., Zhan, J., Althoff, T., McDuff, D., & Liu, X. (2026). Transforming wearable data into personal health insights using large language model agents. Nature Communications, 17, 1143.

2. Ghadi, Y. Y., Shah, S. F. A., Waheed, W., Mazhar, T., Ahmad, W., Saeed, M. M., & Hamam, H. (2025). Integration of wearable technology and artificial intelligence in digital health for remote patient care. Journal of Cloud Computing, 14, 39.

3. OECD. (2026). Building people-centred digital health systems: Lessons from PaRIS. OECD Publishing, Paris.

4. Rebitschek, F. G., Carella, A., Kohlrausch-Pazin, S., Zitzmann, M., Steckelberg, A., & Wilhelm, C. (2025). Evaluating evidence-based health information from generative AI using a cross-sectional study with laypeople seeking screening information. npj Digital Medicine, 8, 343.

5. Shekar, S., Pataranutaporn, P., Sarabu, C., & Cecchi, G. A. (2025). People overtrust AI-generated medical advice despite low accuracy. NEJM AI, 2(6).

6. Bracic, A., Spector-Bagdady, K., Towle, S., Zhang, R., James, C. A., & Price, W. N., II. (2026). Factors for patient trust and acceptance of medical artificial intelligence. JAMA Network Open, 9(3), e260815.

7. Campos, H., & Salmi, L. (2025). Critical AI health literacy as liberation technology: A new skill for patient empowerment. NAM Perspectives.

8. van Kolfschooten, H., & van Oirschot, J. (2024). The EU Artificial Intelligence Act (2024): Implications for healthcare. Health Policy, 149, 105152.

9. Khasentino, J., Belyaeva, A., Liu, X., et al. (2025). A personal health large language model for sleep and fitness coaching. Nature Medicine, 31, 3394–3403.

10. Ankolekar, A. (2026). The right to understand in health care AI. Journal of Medical Internet Research, 28, e95090.

Acknowledgements

Generative-AI tool use: 

Generative AI tools were used solely for language editing and text refinement. The ideas and perspectives presented in this blog are the author’s own and are informed by the references cited in the text.

Disclaimer

This blog is a contribution of expertise from our volunteers. It is not a reflection of any opinion or roadmap of their employers. All blogs from WAI Labs go through a review and/or editing as needed, and are vetted for veracity. For questions or comments, write to wailabs@womeninai.co .