First Google, now AI. Will doctors hate us?

It begins, as many things do, in the dark. A cough, a pinch of pain, a sudden awareness of the body’s fragility at two in the morning. Once, we would have reached for Google, typing in “heart attack?” or “what does chest pain mean,” and sifting through the chaos of WebMD. Now the hand moves differently. It unlocks a phone, calls up Bing, Copilot, or Ada. The machine does not simply list ailments. It interrogates. It asks, and asks again, in a patient monotone, until finally it offers a diagnosis with unsettling confidence.

Every day, more than fifty million such encounters occur across Microsoft’s AI systems. Fifty million people, half-believing, half-fearing the verdict of an artificial intelligence. Fifty million people, in bedrooms and kitchens and cars, wondering if they should trust the ghost in the machine.

The End of Gatekeepers

Once, doctors held the keys. They were the translators of pain into language, the custodians of knowledge. We arrived at their offices with nothing more than our bodies, waiting to be told what they meant. At most, we carried a sheaf of internet printouts, nervously underlined, embarrassing to us and faintly insulting to them.

But now the gate has been left swinging. AI does not blush, does not sigh, does not tell us to wait three weeks for an appointment. It processes more medical literature in a minute than a doctor could in a lifetime. Microsoft’s own “Diagnostic Orchestrator” managed to identify the correct illness in 85.5 percent of famously complex New England Journal of Medicine cases. The physicians, left unaided, managed twenty.

It is possible, of course, to quarrel with these numbers, to say the cases were designed for machines, to say the doctors were hobbled. Still, the fact remains: the aura of professional monopoly has cracked. We sense it. They sense it. And they sure don’t like it.

The Digital Doctor at Home

Picture a man with heartburn after dinner, hesitating before the sink. He opens Ada. Twelve brisk questions later, the machine pronounces: gastroesophageal reflux. A strange comfort settles. This scene repeats fifteen million times a month. Fifteen million! And unlike the crude symptom checkers of a decade ago, these new tools learn. They speak in natural language, recall demographic detail, and even adjust for local disease prevalence. They resemble less the frantic “could it be cancer?” searches of early Web 1.0 and more the calm blue dot of GPS: locating us precisely, telling us where we stand. And like GPS, they improve as we use them. The more patients confide, the sharper the machine becomes. In this way, the relationship deepens: our data feeds it, and its answers shape us. And it is not only the body that people bring to these encounters. Increasingly, they turn to AI to soothe their fears, to confess sleepless worries, to seek mental support once reserved for spouses or friends, even seeking a different type of virtual relationships through hot AI chat platforms.

When Patients Arrive Pre-Diagnosed

A Monday morning in the clinic. One patient swears she suffers from autoimmune disease, though her watery eyes betray only pollen. Another confesses he delayed treatment for chest pain, persuaded by AI that it was “probably anxiety.” A third has correctly identified a skin condition and recites treatment options better than a trainee physician.

This is not fantasy. This is ordinary life now. The consultation has become a negotiation between three parties: patient, doctor, and algorithm. And the old asymmetry of knowledge has flattened.

Two-thirds of doctors report using AI tools themselves, yet many are unsettled. The authority they once carried into every encounter now meets competition, sometimes from the very people they are meant to guide. A consultation can dissolve into a debate over probabilities. And beneath it all lies the uneasy question of blame. If an AI makes a mistake and a patient suffers, who is responsible? The doctor who never gave advice? The company whose code cannot be cross-examined? Or the patient who trusted too quickly?

Accuracy and the Uneasy Comparison

The numbers are mixed. Some studies suggest an accuracy of about 35%, for the general AI tools like ChatGPT. Numbers from the previously mentioned Diagnostic Orchestrator are obviously higher, as it is the specialized AI model. Still, the figures unsettle. Machines make different mistakes. They are strong on rare conditions, weak on intuition. They recognize patterns across oceans of data, yet cannot see the small tremor in a hand, the hesitation before an answer. They miss the human residue.

And yet, we know this too: doctors make mistakes as well. Only their errors come wrapped in narrative, explanations, hesitations, the saving grace of “something doesn’t fit.” AI has no such second thoughts. It is wrong, sometimes disastrously wrong, but wrong with confidence.

This is the unease: that we might believe it more readily than we believe ourselves.

Behind all these questions of pride and authority lurks another, more merciless truth: money. Health care in the United States devours nearly a fifth of GDP. A single visit to a primary care physician costs hundreds of dollars, before tests and referrals swell the bill.

Now imagine. AI gives you an answer at midnight, for free. Imagine insurers insisting you consult an algorithm before they cover a doctor. Imagine telemedicine platforms blending AI triage with human oversight, selling the package for a fraction of the old price. The math is relentless. If the machine can rule out heartburn without blood tests, without follow-ups, without co-pays, why fund the old apparatus at all? And what, then, of medical education? Why invest a decade of training and oceans of debt in cultivating diagnostic reasoning, if diagnostic reasoning is precisely what machines excel at? The incentives lean inexorably toward automation, and doctors feel the economic ground shift beneath them.

A Prescription for Coexistence

Yet we should resist simple apocalypse. The physician will not vanish. Something else will happen. Machines excel at what is enumerable: probabilities, data, and symptoms parsed into categories. Humans excel at what is inexpressible: hesitation, intuition, the reading of silences. AI may tell you reflux, but it cannot hear the way you describe your marriage while mentioning reflux. It cannot sense the grief behind the sleeplessness. It cannot, finally, place a hand on your arm.

The most plausible future is not elimination but collaboration. AI to sweep away the noise, doctors to listen for what is unsaid. Machines to organize the mess, humans to confront the mystery.

Doctors have adapted before. They accepted X-rays, blood tests, and electronic records. They even tolerated Google in the waiting room. They will adapt again. But only if they admit the monopoly is over and treat this not as an insult but as an invitation. Already, sixty-eight percent of physicians concede that AI offers genuine advantages. The task is to weave those advantages into practice without unraveling what makes medicine human.

We should not say, then, that doctors will hate us for using AI. Instead, they will have to decide whether to share authority or to lose it altogether.