Radiologists Thrive Despite AI Predictions, Salaries Hit $571K

In 2016, Geoffrey Hinton—widely known as the "godfather of AI"—made a stark prediction: people should stop training new radiologists because artificial intelligence would soon outperform them at reading medical images. He gave the profession five to ten years before obsolescence.

Nearly a decade later, radiology tells a different story. The number of active radiologists in the United States has grown roughly 10% over the past ten years, according to Christoph Herpfer, an economist at the University of Virginia's Darden School of Business who studies physician labor markets. Rather than vanishing, the field now faces a significant shortage.

As of May, there were 7,469 active radiology job postings nationwide, with 1,470 remaining unfilled for more than 60 days, according to RadBoard.io, a radiology-focused recruiting platform. Average radiologist salaries have climbed to $571,000 in 2025, up 9% year-over-year, per Medscape data.

Why it matters

Radiology's resilience offers a concrete case study for business leaders navigating AI adoption. The field demonstrates that even when AI automates specific tasks effectively, complex professional roles often expand rather than disappear—a pattern with implications across industries from accounting to legal services.

What changed the forecast

Hinton himself walked back his prediction last year, clarifying he was speaking narrowly about image analysis rather than the entire profession, as reported by the New York Times. He now envisions radiologists working alongside AI for greater efficiency.

Several structural factors have protected the profession. Medicare and Medicaid only reimburse radiology studies when a licensed physician performs the final interpretation. Legal liability for missed diagnoses remains unresolved for AI systems. And reading scans represents just one component of radiologists' work—they also consult with other physicians, monitor patients, and in some specializations perform hands-on procedures.

"Complex jobs like being a doctor consist of many sub-tasks," Herpfer explained to Fortune. "Even if you can automate one or two of those, you just expand the time you spend on the other tasks."

Demand has also intensified. Between 2018 and early 2025, radiology case loads jumped 25%, according to the Journal of the American College of Radiology. Ironically, FDA-approved AI tools that make imaging faster and cheaper to produce have contributed to this surge.

Nvidia CEO Jensen Huang and Netflix cofounder Reed Hastings have both recently pointed to radiology as evidence that AI doomers conflate task automation with job elimination.

The human element persists

Dr. Jeff Chang, a former ER radiologist who cofounded AI health care startup Rad AI in 2018, spent a decade reading 150 to 200 imaging studies per night shift. His company's AI tools now save radiologists close to an hour per shift by automatically generating report conclusions—but he doesn't believe the technology can fully replace the profession.

Dr. Tonie Reincke, a Texas-based interventional radiologist, emphasized the irreplaceable human aspects: "A computer can't hold a patient's hand when they're crying. A computer can't hand them a tissue."

Herpfer drew parallels to accountants, who were expected to be eliminated by spreadsheet software in the 1990s. Instead, Excel removed routine number-crunching and freed accountants for more complex advisory work.

"As long as AI doesn't make this quantum leap of becoming sort of AGI, as long as this extreme scenario doesn't happen, most jobs in the medium run are probably going to be reasonably safe," Herpfer said. "That's the lesson I think we can learn from the radiologists."

These details were first reported by Fortune.