One in four Americans have used AI for health information or advice. That number, from the West Health-Gallup Center on Healthcare in America survey of more than 5,500 U.S. adults, is striking. But for healthcare system leaders, the data illuminates something larger than a technology adoption trend. They reflect where patient expectations have moved and where the healthcare system has not yet caught up.
The tools are new. The need is not.
Patients have sought health information outside the exam room for decades. This includes policy shifts such as the FDA’s relaxation of direct-to-consumer drug advertising in 1997, as well as the rise of tools such as WebMD, Google symptom searches and patient portals that give access to lab results before a provider calls. Each iteration reflected the same underlying reality - people want information on their own terms, and they will find it.
AI is the latest and perhaps the most powerful expression of that drive. What has changed is not the behavior, but the capability. Patients are no longer just looking things up. They are having conversations, interpreting their own records and arriving at clinical encounters having already done significant preparatory work. The system that ignores this dynamic does not stop it. It simply steps out of it.
AI is exposing gaps the system has not closed.
The data also reveals something important about why people are turning to AI. Among recent users, 16% did so because they could not access a provider, and 14% because they could not afford a visit. These are not patients choosing convenience. They are patients who encountered a barrier and found the next available option.
Income tells the starkest story. Among adults earning less than $24,000 annually, 32% report using AI because they could not pay for a doctor’s visit, compared with 2% of those earning $180,000 or more. AI, in these cases, is functioning as a safety valve for a system under pressure.
For healthcare organizations serving Medicaid-heavy populations, rural communities or those with explicit community benefit commitments, that pressure is most acute. When healthcare is out of reach, patients may be more likely to turn to AI instead.
A more informed patient is changing the clinical relationship.
Most patients using AI are not trying to bypass their providers. Nearly 60% of recent AI users say they research on their own before seeing a doctor. Nearly half, 46%, say AI made them feel more confident when talking with or asking questions of a provider. Twenty-two percent say it helped them identify issues earlier.
These are the behaviors of patients who want to engage more fully with their own care. When someone arrives having already reviewed their condition, formulated questions and worked through their concerns, the patient-provider encounter changes. The dynamic shifts from the clinician as sole authority toward something closer to a collaboration, and that shift, when supported well, can lead to better outcomes, higher adherence and stronger therapeutic relationships.
Healthcare organizations including Sutter Health are already integrating AI directly into the patient experience, embedding tools like Epic's Emmie within its existing patient portal so patients can ask questions, prepare for visits, and engage with their own records in plain language.
That shift creates both a challenge and an opening. Providers that help patients navigate toward reliable information and serve as the trusted guide for interpreting what they find, including AI-generated insights shaping their care decisions, become a more essential partner in their care, and a more trusted one.
Trust is the real tension.
Here is the friction point: only 4% of recent AI users say they strongly trust the accuracy of AI-generated health information. Yet they are using it anyway. This is not an argument for alarm. It is a clinical reality that healthcare organizations are uniquely positioned to address.
AI tools are imperfect. They can misinterpret symptoms, miss critical context and are not a substitute for a clinical diagnosis or a conversation with a provider who knows a patient’s history. Patients, by and large, know this and are using these tools anyway, often because the alternative is waiting, going without or navigating a system that is not available when they need it. They want to better understand their health, ask informed questions, and advocate for their care.
That dynamic matters for how healthcare organizations respond. The goal is not to compete with AI-generated information or dismiss what patients find. It is to be the trusted layer on top of it - the clinical relationship that contextualizes, corrects where needed, and affirms where appropriate. Clinicians who invite open dialogue about what patients have read create the conditions for that trust to deepen. Those who treat patient research as noise to be overridden risk losing the relationship entirely to tools that, for all their limitations, are available whenever patients need them.
What leading organizations will do differently.
The healthcare organizations that will define this next chapter are not waiting for AI to be perfected or regulated into a tidy framework. They are making deliberate choices now about where they show up in the patient journey. Three choices will distinguish the organizations that will successfully adapt to this new era from those that will follow:
Be the trusted source throughout the patient journey.
Patients are already researching, preparing and forming expectations before they arrive. Organizations that proactively connect patients to trusted, EHR-embedded AI tools shape the quality of that preparation in ways that general-purpose consumer platforms cannot. The clinical encounter becomes more productive, and effectively longer, because the health system is guiding the patient in learning about their health both before and after the time spent with a clinician. After the visit, that same trusted access helps patients understand what was covered, answer lingering questions, and stay engaged in their ongoing care.
Prepare clinical teams for a different kind of patient.
The shift from authority to collaboration does not happen automatically. Clinicians who are prepared to acknowledge what patients have found, engage with their questions and build on their existing understanding will drive stronger outcomes than those who treat patient research as a distraction. That preparation is a leadership and culture investment, not a technology one.
Treat equity as a strategic signal, not a footnote.
Lower-income patients are turning to AI in disproportionate numbers because cost and access barriers leave them few alternatives. Healthcare organizations already committed to those communities have a distinctive opportunity to ensure AI becomes a bridge to better care - one that connects rather than substitutes.
West Health and the American Hospital Association have launched a joint initiative to help healthcare organizations act on exactly these opportunities - sharing technologies, frameworks and best practices for integrating AI into patient care safely and effectively. Because the question for healthcare system leaders is no longer whether AI will be part of how your patients experience healthcare. It already is. The only remaining question is whether your organization will help lead what that looks like.
About the Authors
Melissa Martin is executive director of HC65 at West Health, where she leads a national initiative focused on advancing technology-enabled care delivery. She brings experience across the healthcare ecosystem, including pharmaceutical companies, large health systems, federal healthcare, and consulting, with deep expertise in EHR transformation, applied AI, and care delivery redesign. Melissa holds a Master of Health Administration and began her career as a television news journalist before a personal experience navigating the healthcare system with a family member inspired her to dedicate her career to improving how care is accessed, delivered, and experienced.
Kevin Biese, MD, is a board member of the West Health Policy Center and associate professor of Emergency Medicine and Internal Medicine at the University of North Carolina at Chapel Hill School of Medicine, where he also serves as vice chair of Academic Affairs and director of the Division of Geriatric Emergency Medicine. He is co-leader of the national Geriatric Emergency Department Collaborative and chairs the board of governors for the American College of Emergency Physicians Geriatric Emergency Department Accreditation Program, which has improved the quality of care in more than 475 emergency departments across 40 states and four countries.