Doctors and AI: what 271 million clinical questions revealed

Artificial intelligence is used in clinical medicine in two main ways: it assesses test results on its own, above all medical images, and it helps doctors find the information they need. AI is also used in healthcare management and in research.

Where AI reads scans

How well AI assesses images depends on the volume and quality of the data it was trained on. That is why it performs best where a great many examinations are done, such as routine chest X-rays or screening mammography. Differences between patient groups, such as sex or race, limit it. AI also reaches its conclusions differently from a human, which can confuse the doctor who is responsible for the final report.

On the other hand, AI takes routine tasks off the doctor's plate and flags abnormalities that are hard to spot with the naked eye. The models in use increase the sensitivity of examinations, but questions remain about how well their conclusions carry over to other data.

What the OpenEvidence study showed

Another use is growing fast: getting professional information in real time. A doctor asks a question in plain language, and the system is meant to back up its answer with a reference to a reliable source in the scientific literature, where it can be checked. This is what sets a doctor's exchange with a clinical system apart from an ordinary internet search. A major advantage is the ability to discuss a specific patient's treatment plan with AI, taking into account their illnesses, general condition and the interactions between all the medications they take.

According to a study published in JAMA Network, from 1 January 2024 to 30 June 2026 the free clinical platform OpenEvidence was used by 950,000 people, who asked a total of 271 million questions. More than half of the users, 550,000, were doctors; they asked an average of 37 questions a month, most often about treatment, diagnosis and patient management. The remaining users were nurses, physician assistants and pharmacists. The number of users grew markedly over the period.

How medical learning is changing

Doctors with less experience asked questions more often. This may point not only to less knowledge but also to a generational shift in how information is obtained. A notable share of the questions concerned areas outside the doctor's own specialty. The results hint that the way knowledge is acquired is changing too: doctors used to add to it mainly through training and formal education, and now increasingly through targeted searches for specific facts.

The body of medical knowledge is growing exponentially and long ago stopped fitting in one person's head, which is why medicine has split into so many specialties. AI partly replaces the labor-intensive medical consultation between colleagues. It cannot yet fully replace the doctor responsible for diagnosis and treatment, but it can already offer very useful, high-quality professional support.

Around Prague, based on reporting by osel.cz