As generative AI tools increasingly shape how patients research symptoms, diagnoses and treatments, chief medical information officers say the challenge is not eliminating AI from clinical conversations — but helping clinicians respond with clarity, consistency and trust.
Across health systems, CMIOs describe three ways they are guiding clinicians as patient-generated AI becomes a routine part of care encounters.
Reframing AI as a starting point — not a diagnosis
Several CMIOs say clinicians are encouraged to acknowledge AI-generated content respectfully and use it as an entry point for discussion, rather than a conclusion to be validated or dismissed.
At University of California Irvine Health, CMIO Deepti Pandita, MD, said clinicians are trained to gently correct misinformation and reposition AI outputs as tools for education and shared decision-making. The organization reinforces a “human in the loop” message to clarify that AI is not medical advice and is investing in clinician training around AI literacy.
That framing mirrors how some leaders view AI in the broader history of patient-sourced information. At Los Angeles-based Mission Community Hospital, Jason La Marca, MD, CMIO, said clinicians are encouraged to treat AI-generated content as a conversation starter — similar to the WebMD or Medscape printouts patients have long brought into appointments.
“It really isn’t any different,” he said. “We learn with them and they learn from us.”
At Scottsdale, Ariz.-based HonorHealth, CMIO Matthew Anderson, MD, said clinicians approach AI-generated information much like Google searches or social media forums, using it to support discussion and motivational interviewing rather than as a source of clinical authority.
Standardizing how clinicians set boundaries and correct misinformation
Other CMIOs are focusing on consistency — not just in how clinicians acknowledge AI, but in how they explain its limits.
At Virginia Hospital Center in Arlington, CMIO Usman Akhtar, MD, said clinicians are being coached to deliver a brief, standardized message: AI can help patients prepare questions, but it is not a diagnosis. Guidance also helps clinicians set boundaries around what can be safely addressed during a visit versus what requires further evaluation, while keeping documentation grounded in clinical reasoning.
At Roseville, Calif.-based Adventist Health, CMIO Amer Saati, MD, said clinicians are instructed to acknowledge patient-generated AI content, correct misinformation directly and anchor every encounter in evidence-based care. Organization-approved AI tools can support patient education and shared decision-making, he said, but clinical assessment remains central.
Some systems acknowledge they are still early in developing formal guidance. At Wenatchee, Wash.-based Confluence Health, CMIO Becket Mahnke, MD, said the organization does not yet have system-level policies for managing patient-generated AI content. However, after a recent Wall Street Journal article prompted patient questions about ambient documentation, the system proactively provided clinicians with scripted responses so they would be prepared.
Building accountability into AI tools themselves
Beyond communication strategies, some CMIOs say accountability must be embedded directly into AI workflows.
At Fort Wayne, Ind.-based Parkview Health, CMIO Mark Mabus, MD, described the organization’s approach as “trust but verify.” Generative AI tools include built-in feedback mechanisms that allow clinicians to rate outputs at the point of care. That feedback is tracked internally and shared with vendors to refine the tools.
“If a clinician uses the content provided by AI, they understand that they are responsible for the generated output,” he said.
The shift, leaders say, is less about controlling what patients bring into appointments and more about equipping clinicians with the language, guardrails and accountability needed to keep AI-informed conversations grounded in safe, evidence-based care.
