AI Is Now Catching Doctor Errors

David Gewirtz, Senior Contributing Editor at ZDNET, published his account on May 27, 2026. He visited a trained ophthalmologist — not just an optometrist, but a medical doctor specializing in eye disease and surgery. The doctor got his distance prescription right. The computer glasses prescription? Completely wrong distance calculation.

Gewirtz didn't just complain about it. He ran the numbers through ChatGPT, Claude, and Gemini.

All three AI systems agreed. The doctor had written the prescription for the wrong focal distance. Gewirtz's monitor sits 23 inches from his eyes — stretching to 29 inches at the edges of his 38-inch curved screen. The prescription didn't account for that. The AI-derived glasses worked. The doctor's original pair failed.

What the AI Actually Did

This wasn't AI playing doctor. It was AI doing math and cross-referencing optics — exactly the kind of systematic, detail-oriented work that gets rushed in a 15-minute exam room visit.

According to Gewirtz's account in ZDNET, the eye chart tests that determine prescriptions are highly subjective — rapid lens swaps under less-than-ideal conditions, with patients second-guessing every choice. High stakes, limited time, imperfect conditions.

Three different AI platforms independently caught what a human specialist missed.

What the Broader Research Shows

This pattern extends beyond one case. According to a 2021 review published in Future Healthcare Journal — authored by researchers at Microsoft Research Cambridge and University College London, including Dr. Junaid Bajwa and Professor Bryan Williams — AI has the potential to "fundamentally transform the practice of medicine and the delivery of healthcare."

They identified the core problem: healthcare systems globally are straining under aging populations, chronic disease burdens, and rising costs. AI isn't a luxury addition. It's increasingly a structural necessity.

Built In, updated in March 2026, reports that tools like ChatGPT Health and Claude for Life Sciences are now enabling users to connect wearable device data directly into AI models for real-time, personalized health recommendations. Companies like Pfizer, Butterfly Network, and Tempus are already deploying AI to speed up diagnostics, cut costs, and personalize treatment.

AI Multiple's healthcare use case database documents at least 25 distinct clinical and administrative applications already in active deployment — from drug discovery to robotic surgery assistance.

Coverage Gaps

Most outlets covering AI in healthcare fall into predictable categories.

The center-left press, including outlets like MIT Technology Review, frames AI healthcare stories primarily through an ethics and access lens — important questions, but they often overshadow the basic factual story: AI is already outperforming doctors in specific, documented cases.

The trade and tech press, like Built In and ZDNET, gets closer to the ground truth but tends to bury the implications. Gewirtz's piece is significant but published in a personal blog format, limiting its reach.

When three AI systems catch an error a specialist missed, that's a medical accountability story.

The Practical Question

How many people are walking around with wrong prescriptions — glasses, medications, diagnoses — because a rushed doctor made a calculation error that a free AI tool would have caught in 30 seconds?

No one is tracking this systematically.

The medical establishment has institutional incentives to frame AI as a "supplement" to doctor judgment, never a check on it. Medical licensing boards, malpractice insurers, hospital systems — none of them benefit from a public conversation about AI catching doctor errors at scale.

But the documented cases keep accumulating.

What This Means for Patients

If your doctor gives you a prescription — glasses, medication, anything quantitative — and something feels off, you now have tools to check the math. ChatGPT, Claude, and Gemini are free or near-free. They don't get tired. They don't rush through appointments. They don't have a waiting room full of patients.

This isn't about replacing your doctor. It's about being an informed patient in a system that is structurally overloaded and demonstrably fallible.

Use the tools. Ask the questions.