Alan P. Feren, MD
AI can help organize your health story. It should not replace the clinical work needed to understand it.
Perhaps you have asked an artificial intelligence tool to help explain a medical term or organize symptoms before an appointment. Perhaps your clinician uses one to prepare notes or summarize your medical record.
These tools may be useful. But a well-organized account of your symptoms is not the same as an understanding of what is happening to you.
The American Medical Association’s 2026 physician survey found that 81 percent of respondents used AI professionally, up from 38 percent in 2023. Those uses include summarizing research and preparing clinical documentation—not simply diagnosing patients. (American Medical Association)
For patients and families, the important question is not just whether AI is involved. It is whether someone is still listening carefully, examining appropriately, questioning the first explanation, and taking responsibility for the treatment plan.
The intelligence you should expect from your clinician
I call that work Actual Intelligence: the clinician’s trained ability to listen, observe, examine, consider possible explanations, recognize uncertainty, and decide what should happen next.
It is not intuition alone. It is not a belief that people are always right and computers are always wrong. It is the careful clinical work that gives information meaning.
You experience it when your clinician asks what changed rather than assuming a familiar symptom has a familiar cause. You experience it when reassuring test results are explained in relation to your concern, rather than offered as the end of the conversation. And you experience it when a treatment recommendation accounts for the medications you already take, the help available at home, and what you can realistically manage.
AI can assist with information. The clinician remains responsible for deciding how that information applies to you.
A convincing answer can still be incomplete
AI’s mistakes are not always obvious. Its explanations can sound calm, clear, and convincing while leaving out something important. The World Health Organization has warned that these tools can produce incomplete or inaccurate health information and encourage people—including health professionals—to overlook problems in an automated answer. (World Health Organization)
That is why a polished summary deserves a question: Does this faithfully represent the person, or does it merely sound reasonable?
Consider a hypothetical patient who has been experiencing dizziness and a heaviness that sleep does not relieve. He struggles to describe it, so he uses an AI tool to organize his symptoms before an appointment. He is looking for language, not a diagnosis.
His laboratory results are reassuring. During the visit, he says, “It still doesn’t feel right. I used an AI tool, and it made me wonder…”
The clinician explains that AI does not have the full clinical picture. That is reasonable advice. But the conversation moves on without exploring what still concerns him.
The missing step is not automatically another test. It is a question:
“Tell me what still feels wrong.”
The AI summary may have accurately organized what the patient entered. The clinician’s note may accurately record what was discussed. Neither establishes that the underlying concern was understood.
The patient’s unease does not prove that a serious illness has been missed. It does mean there is something left to explore.
Your experience should remain the starting point
There is nothing wrong with seeking help to describe a confusing symptom. The important distinction is between using AI to express your experience and allowing its explanation to replace your experience.
Before asking a tool to organize your story, try putting the concern into your own words: what changed, when it began, how it affects your day, and what worries you. The language does not need to sound medical.
When you review the summary, consider whether it still reflects what you meant. Did it leave something out? Did it make an uncertain observation sound definite? Did it replace your main concern with a different one?
You can bring that distinction into the appointment:
“AI helped me organize this, but here is what I actually noticed—and what still concerns me.”
Or:
“The summary sounds clear, but it leaves out the part I most wanted to discuss.”
A family member or caregiver can help you prepare or remember questions. Their observations can contribute to the conversation without displacing your own account.
These are ways to participate, not requirements you must fulfill to deserve careful attention. You do not need to use AI, arrive with a polished summary, or know which medical questions to ask. Helping you develop the story is part of clinical care.
Why clinicians need to keep practicing their judgment
The same AMA survey found that 88 percent of physicians expressed at least some concern about AI-related skill loss. That measures concern; it does not establish that those physicians have lost their skills. (American Medical Association)
The concern is nevertheless important. When a tool supplies the first explanation repeatedly, a clinician may spend more time editing that explanation and less time independently working through the possibilities. The question becomes whether a useful assistant is gradually taking over work the clinician needs to keep practicing.
For patients, this is not a reason to distrust every clinician who uses AI. It is a reason to value the thinking that should remain visible in the encounter: asking further questions, considering what does not fit, and explaining why a recommendation makes sense.
If we expect clinicians to recognize when AI is misleading, their ability to make that judgment must remain actively practiced.
Patients should not be responsible for testing their clinician’s competence or detecting every AI error. Maintaining clinical skills and overseeing the technology belong to the professionals and organizations providing care.
What “Actual Intelligence first” means for you
AI may help your clinician review information or prepare documentation. The safeguard is that the clinical assessment guides how its output is used—not that a ready-made answer becomes the assessment without sufficient questioning.
You can ask:
“How does this explanation fit what I have told you?”
“What remains uncertain?”
“What would make us reconsider the treatment plan?”
The purpose is not to challenge every recommendation. It is to understand the reasoning well enough to participate in your care.
A clinician may appropriately disagree with an AI suggestion. Explaining why it does not fit your history or findings can be more useful than either accepting it automatically or dismissing it without discussion.
Likewise, an AI-assisted note still needs clinical review. A signature should represent responsibility for the account, not merely confirmation that the wording reads smoothly. WHO’s guidance emphasizes human oversight and accountability in the use of AI for health. (World Health Organization)
A useful answer is not yet actual care
This is part of what I mean by Care-Full Medicine. Actual care involves an explanation you understand, a treatment plan you can carry out, and clarity about what to do when something changes.
A recommendation may look appropriate on a screen while overlooking that you cannot afford it, cannot travel to the appointment, or need help managing it alongside other treatments. Those realities belong in the clinical conversation.
AI should support that conversation, not make it unnecessary.
The goal is not a choice between modern technology and attentive medicine. It is to use technology in ways that preserve the listening, examination, reasoning, and Follow-Through on which care depends.
You do not need your clinician to compete with a computer. You need your clinician to understand you—and remain responsible for the decisions made with you.
That is why Actual Intelligence must remain first.
Adapted for Amazing Care Network from “Actual Intelligence: the skill AI cannot replace,” originally published on KevinMD. (KevinMD) on 29 September 2026
