AI Won't Replace Radiologists, But Their Job Is Changing Fast
A bold 2016 prediction that computers would wipe out radiology within five years turned out to be wrong. What actually happened is more interesting.
Key points
- As of early 2026, roughly three-quarters of the 1,400 AI-enabled medical devices cleared by the US Food and Drug Administration are designed for radiology.
- Geoffrey Hinton, the Nobel Prize-winning researcher often called the "godfather of AI", predicted in 2016 that radiologists would be replaced by computers within five years.
- The number of practising radiologists is expected to grow by 26 percent or more over the next 30 years.
- A review of 43 clinical trials found that AI-assisted colonoscopies detect more polyps than conventional ones.
- AI tools in radiology now match or beat human performance on specific tasks, even as the profession itself keeps expanding.
Geoffrey Hinton made a bold call in 2016. The Nobel Prize-winning researcher, widely known as the "godfather of AI", said radiologists, the doctors who read X-rays and other medical images to help diagnose illness, would be out of a job within five years, replaced by computers. We've covered Hinton's evolving views on the technology he helped build in our August reporting on his regrets about modern AI.
He was wrong about the job losses. Right about almost everything else.
Radiologists aren't disappearing. They're projected to grow in number by 26 percent or more over the next three decades. The field's response to Hinton might borrow from Mark Twain: the report of its death was an exaggeration.
So what did AI actually do to radiology?
It moved in as a colleague, not a replacement. Radiology is, by a wide margin, the busiest corner of medicine for AI tools right now. As of early 2026, about three-quarters of the 1,400 AI-enabled medical devices cleared by the Food and Drug Administration were built for radiology.
Some of those tools handle the admin side: drafting reports, flagging the scans that need a doctor's eyes most urgently. Useful, but not dramatic.
Other tools are something else entirely. They can spot abnormalities invisible to the human eye, and on certain tasks they match trained radiologists or exceed them. A review of 43 clinical trials found that AI-assisted colonoscopies detect more polyps than traditional ones, a genuinely significant finding for anyone worried about bowel cancer.
What does this mean for patients?
In the short term, mostly good news. AI's catching things that might otherwise be missed and helping doctors work through scans faster.
The honest caveat: these tools are still aids, not autonomous doctors. A radiologist's still reading your scan. The AI is nudging them toward things worth a second look.
Privacy is worth a mention. Medical imaging AI learns from patient scans. If you want to know whether your hospital uses AI tools and how your data's handled, that's a fair question to put to your care team.
What happens next?
Radiology's a preview of something bigger. If AI can work alongside expert humans in one of medicine's most pattern-recognition-heavy fields, the same shift will follow elsewhere in healthcare.
Hinton was off on the timeline and wrong about mass displacement. But his core point, that machines would reach human-level performance on reading medical images, has arrived. The profession adapted. That story's still playing out, and if the early-stage brain-scan AI we reported on in July is any guide, radiology won't be the last specialty reshaped by it.
Common questions
Will AI replace my radiologist soon?
No. The number of radiologists is growing, not shrinking. AI tools assist doctors rather than replace them, flagging urgent scans and catching hard-to-see details, but a qualified physician is still responsible for your diagnosis.
Is my medical imaging data used to train AI?
It can be, depending on your hospital and which tools they use. Ask your provider for their data-use policy if that concerns you; it's your right to know.
