Patients value disclosure and want opportunities to consent to AI use in their care. Patients want to learn about AI tools where meaningful patient awareness and consent are feasible, regardless of whether they interact with them directly. Patients said they wanted to receive this information through multiple communication channels — such as being told during an appointment as well as receiving written notice beforehand — so they can digest the information on their own time and have the chance to ask questions. “I would want them to be like, ‘Hey, listen, I have this new tool. Let me just explain what it does, and how it’s here to help you,” said Jessica in The Bronx, New York. “I’d rather you talk to me about it first before you just stick me with something and I don’t know what’s happening.”
Patients described being informed about a “policy change at an institution,” while another said they want “something like HIPAA for generative AI.” Another patient said, “I always tell people [AI is] here, but somebody’s still got to take care of it. Somebody got to run it,” pointing toward a desire for management and accountability.
Patients want AI to create more opportunities for accountability and accuracy in their care. Patients appreciate the possibility of AI tools creating plain-language notes — or “paper trails” — on their doctor visits so they can verify information, dispute inaccuracies, prepare questions, and advocate for their health needs — potentially reflecting gaps patients experience in their primary care. “I think it [technology in health care] keeps people more up to date of what was actually said, what was done,” said Stephanie from Pulaski, Virginia. “Because in the past, before that, if I left the doctor’s office, whatever was said there and what I remembered or wrote down was all I knew. And now I can go on a portal or something and see what’s there.”
Discussion
Health systems, clinicians, and patients are turning to AI tools to address issues facing primary care. Our interviews revealed that patients want transparency about when and why AI tools are used in their care, opportunities to ask questions, and — when possible — the ability to decline without compromising the quality of their care.
Their views were shaped by whether they trusted their clinicians, previous negative experiences in health care, and everyday experiences with technology. Patients expressed hope that AI could improve care and communication, as well as concern that it could erode clinical judgment, privacy, and the human connection central to primary care. Some saw AI’s growth as inevitable and largely outside of their control, while others were actively embracing it to better understand and manage their care.
Across these perspectives, patients consistently wanted transparency and accountability, meaningful benefits, and choice where possible. Below, we highlight opportunities for clinicians and health system leaders to make improvements based on these themes.
Transparency and Consent
Because AI can shape primary care through its role in documentation, patient communication, and clinical decision support, patients viewed transparency about AI — including benefits and risks — as a prerequisite for trust and responsible use. Other research points in the same direction: most U.S. adults say they want to be notified when AI is used in their health care, and many report limited trust that health systems will use AI responsibly.
Patients described consent as a conversation with their clinician before an AI tool is used, paired with plain-language written information they can review and return to later. They wanted time to think and ask questions, and the option to decline when feasible, and clarity when it isn’t. When real choice isn’t possible, patients prefer disclosure over hollow consent processes.
Recommendations: Clinicians and practice staff could explain the parameters of their AI use to patients through portal messages prior to the visit or plain-language handouts at the visit. Health systems could set rules for when patient consent is required versus when disclosure is enough. When clinically and operationally possible, practices could offer patients a meaningful opportunity to opt out of AI-enabled functions, such as automated note drafting, patient-facing messages, or algorithmic decision support, and offer a non-AI pathway that does not change the quality of care they receive. When opting out is not possible, practices can be transparent with patients about the benefits and risks associated with the tool.
Accountability
Patients expect their primary care clinicians to be responsible for their care, and this extends to AI use as well. Survey research suggests people are more comfortable with AI use when clinicians stay meaningfully involved in care and health systems have strong governance. Patients should be able to tell who makes the final clinical decision — their clinician or an AI tool — and who is responsible for safety and oversight if something goes wrong.
Our interviews reinforced why this matters. Patients wanted AI to assist, not replace, their clinicians’ judgment, and they wanted a clear path to a human who can explain decisions, correct errors, and take responsibility.
Recommendations: One option for health systems is to be clear about who is responsible for AI-supported decisions. Practices could make it easy for patients to reach a human, fix mistakes in their medical records, and appeal decisions that are directly or indirectly influenced by AI. Health systems could also train clinicians and staff to use AI as input — not a shortcut — and monitor for overreliance and harm over time.
Patient Benefit and Fairness
In our interviews, patients pointed to concrete benefits associated with AI tools like clearer visit summaries, plain-language explanations of test results and diagnoses, and easier communication between visits. Some also saw AI as a way to potentially counter dismissive or biased treatment from clinicians and difficulty getting follow-up appointments or second opinions.
At the same time, patients worried that AI could create new barriers for those without reliable internet access, newer devices, or high digital literacy, including some patients in rural communities, who are already underserved. They also worried that any gains in fairness would be undermined if AI tools are inaccurate.
Recommendations: One option for practices is to prioritize AI use cases with clear patient benefit — including better understanding, easier access to care, and preserved human connection — and test whether they reduce patient burden rather than create it. Health systems could measure outcomes that patients notice, like comprehension, perceived burden, trust, and ability to reach a human, alongside safety and efficiency metrics. Additionally, systems could vet tools for equity before rollout, including usability on older devices and with low bandwidth and different literacy levels, as well as monitor for inequitable outcomes. They could also provide nondigital options so that patients are not blocked by the digital divide.
Conclusion
These findings draw on interviews with a diverse group of primary care patients in urban and rural settings with different types of insurance coverage, spanning a range of ages and racial and ethnic identities. Together, their perspectives suggest that the future of AI in primary care will be shaped not only by what these tools can do, but also by whether patients find them understandable, trustworthy, and beneficial in their care.
For health system leaders, successful adoption of AI tools in primary care will depend on their design and whether implementation is done with patients. Primary care clinicians can help ensure that these tools are introduced clearly, used responsibly, and assessed in light of patients’ needs and concerns.
