AI is changing radiology work, but radiologists are not disappearing

In short: AI tools are becoming common in radiology, and they are reshaping how radiologists work instead of replacing them.

What's going on

In 2016, AI scientist Geoffrey Hinton predicted that computers would replace radiologists within five years. Radiologists are still here, and the field is expected to keep growing, with one estimate projecting a 26 percent rise over the next three decades.

AI is now a regular part of imaging medicine. As of early 2026, about three-quarters of the roughly 1,400 AI-enabled medical devices cleared by the US Food and Drug Administration were designed for radiology. Some tools help with speed, like drafting parts of reports or flagging scans that need urgent attention.

Other tools aim to improve accuracy. Researchers estimate error rates for reading medical images at about 3 to 5 percent, which adds up to around 40 million errors worldwide each year. Studies also suggest benefits in nearby areas, like an analysis of 43 clinical trials finding that AI-assisted colonoscopies detected more polyps than conventional ones.

What to watch

Experts say the hard part is teamwork. Many modern AI systems are “black boxes,” meaning they often cannot clearly explain why they made a call (like a calculator that only shows the answer, not the steps). That pushes radiologists into a new role, checking AI results most of the time and catching the rare but important mistakes, while also avoiding trusting the tool too much or dismissing it too quickly.

Training is a key gap. In a 2026 American Medical Association survey, more than a quarter of physicians said they had received no AI training. One common view from radiology leaders is that AI will not replace radiologists, but radiologists who use AI may replace those who do not.

Source: Arstechnica