The news: 72% of consumers say it’s very or extremely important that providers inform them when AI is used in their care, according to a just-published Pew Research Center survey of 3,488 US adults in June. Yet 46% are unclear whether it is being used

Digging into the details: Consumer desire for disclosure increases with the clinical stakes of the AI’s use.

Across clinical and administrative tasks, consumers want providers to explain how AI is used to:

• Analyze their medical scans (81%)
• Make a diagnosis (81%)
• Explain lab results (80%)
• Take notes during a patient appointment (72%)
• Order Rx refills (64%)
• Schedule an appointment (56%)

Most patients don’t just want a quick notice about AI being used in their care—they want more control over the situation.

• 53% say they have little or no say in whether a provider uses AI in their care, while just 17% say they have significant input.
• And three times as many want more control (63%) over how AI is being used in their care as are comfortable with their current level of influence (21%).

Why it matters: As AI rapidly enters patient care, health systems are still deciding how doctors and administrative staff should use it and when to inform patients. Health systems currently use a mix of direct disclosure, patient portal notices, and verbal communication by clinicians to inform patients about AI in their care, though practices vary widely.

Provider organizations are less likely to divulge AI use in routine medical imaging analysis, where radiologists have relied on the technology for years, or in other behind-the-scenes clinical functions than in patient-facing uses such as communicating with patients or explaining test results. At the same time, some health systems have allegedly failed to obtain consent even when AI is being used to transcribe patient-doctor conversations for later clinician review.

Implications for healthcare providers and AI companies: Notifying patients that AI is used in their care isn’t the same as explaining its role in each task and how it may affect their care compared with human-only decision-making. That gap already exists: Just 22% of consumers who say their doctor has used AI in their care have a strong understanding of how it was used, according to Pew.

At the same time, health systems can’t treat every AI interaction the same or give patients equal say over every use. A tiered approach, which could include notification for routine administrative tasks, opt-outs for tools that record or process sensitive information, and explicit discussion for diagnosis or treatment, can give patients meaningful agency without creating consent fatigue. The findings also raise the bar for AI vendors. Tools that make disclosure, human review, and patient opt-outs easy to integrate will be more attractive to health systems seeking to scale AI without damaging the patient relationship.