By Daniel Akst

Once upon a time, my wife's uncle, Jim, delivered babies, set broken bones, diagnosed diseases, and helped people reconcile themselves to mortality. That's what family physicians did in those days.

Things are different now, and the doctor I most often consult is AI. I'd prefer to see Uncle Jim, but if physicians like him still exist somewhere, I doubt I could get an appointment. How I ended up resorting to artificial intelligence -- despite excellent health insurance and proximity to great care -- says a lot about the state of healthcare in this country.

I tried AI out of frustration, necessity, and yes also curiosity, but it has since become my favorite "physician," if you don't mind using the term loosely. Obviously, AI is no substitute for a real doctor in many circumstances. But in many nonemergency circumstances, a real doctor isn't available. And 20 minutes of his time, if you can get it, isn't enough to really know a patient or manage his or her health. My experience hints at the usefulness AI may someday have in finding a way out of our expensive, fragmented and ineffective approach to human health. It also highlights deficiencies in the way even elite care is provided today.

Your time is up

For many readers, the story will be a familiar one. I'm a healthy guy of 69, yet like most aging Americans, I have various low-level chronic disorders. And I get most of my healthcare from a constellation of highly trained specialists, each focused on his or her narrow realm rather than on my overall health. Not only are these tree surgeons blind to the forest, but they are also overwhelmed with patients, which makes appointments scarce and brief. Often, during our fleeting time together, they are more focused on their computers than on me.

My primary-care physician is no more accessible. At my last annual physical I told him I desperately needed somebody to quarterback my care, but that it was nearly impossible to get in to see him. "It's like getting an audience with the pope," he acknowledged. "I can't quarterback anything."

Thus I have some of the best doctors in America, but there is no coherent treatment plan and no very good sense of what to try and fix, what to accept as inevitable, and how to stay healthy. In fact, the system is quite obsessively focused on illness but not much on health. Time constraints all but require an emphasis on handing out prescriptions and pat answers.

When a blood test showed somewhat low iron, for instance, one hurried physician sent me a message saying, "Eat lentils." But I already eat lots of iron-rich red meat. Lentils contain non-heme iron, which is poorly absorbed unless accompanied by food containing vitamin C, all of which the physician left unsaid.

Contrast his terse advice with what I got from Grok, my go-to AI chatbot, when I asked what I ought to eat if my iron levels are a little low. The response, a model of clarity and nuance, was immediate and free: "To raise low iron levels, focus on iron-rich foods and pair them strategically to boost absorption." It then listed foods that might help, starting with meats, followed by three helpful bullet points.

   -- "Pair non-heme iron foods with vitamin C sources (e.g., bell peppers, 
      oranges, tomatoes) to increase absorption." 
 
   -- "Avoid consuming calcium-rich foods (dairy), tea, or coffee with 
      iron-rich meals, as they can inhibit absorption." 
 
   -- "Cook in cast-iron pans to add small amounts of iron to food." 

Ask me anything

It's rare for any doctor to explain this much, pressed as they are or time. But Grok explains everything, and if something seems unclear I can ask a question and get clarification. And ask again. And again. It never runs out of time or patience.

Free testosterone, for instance, can be expressed in several different units of measure. Grok explained this and translated my numbers as needed. I was sent for the test because I was feeling tired. The doctor's web portal showed the results as slightly low. But the reference range was for all men rather than those pushing 70. Also, I was tested in the afternoon, when testosterone is typically lower. Grok said that the test should have been done in the morning, which nobody -- even the lab -- told me. Allowing for my age and the time of day, my results were just fine.

In fairness, Grok depends on me to provide a lot of information I've already gleaned from physicians and reading. And it must be used with caution; AI does make mistakes. "It has made two big ones with me," one acquaintance reported by email, "and when I challenged it, it said, you're right and thanks for calling me out. But they were both kind of worrisome and made me at least somewhat more cautious."

I haven't spotted any errors in my health-related consultations with Grok. And humans make mistakes too; many years ago, a bungled hospital extubation was the likely cause of my grandmother's death.

Never ignore, never forget

On a more quotidian basis, my physicians simply ignore a lot of what I say and, of course, they can't be expected to remember everything about every patient, opening the door to various suboptimal consequences. Grok doesn't ignore or forget anything, and every time it answers a question it does so in light of what I've said or asked before.

Thus it seems able to see the big picture -- and stands as the only one of my medical advisers that recognizes all of me. Recently, after telling Grok of my various diagnoses, I described some mysterious symptoms that any or all of them could be causing (and that my physicians largely ignore). Are these symptoms just an inevitable part of aging, or something worth addressing somehow?

While old age can contribute, Grok opined, "your symptoms are specific and severe enough...to suggest underlying medical issues rather than just aging." Grok then ran through each of the symptoms, with recommended steps for further testing and treatment -- a virtual road map for navigating my health issues. I had never had anything remotely so coherent from any of my doctors, most of whom belong to a healthcare network that could easily use AI to help prepare annual health assessments for patients -- with progress reports, and recommended tests and goals.

I'm blessed to have talented physicians who practice in superb facilities at a world-class medical institution. What I don't have is very good healthcare. AI, the new practitioner in town, is helping me to resolve this paradox. I have no doubt, given physician shortages, soaring healthcare costs and an aging society, that the role of AI will only grow. Based on my own experience, so far so good.

Daniel Akst is a writer in New York's Hudson River Valley. He can be reached at reports@wsj.com.

(END) Dow Jones Newswires

October 29, 2025 19:00 ET (23:00 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.