How are residents using AI?

A new study led by Yale School of Medicine’s Wendy Barr, M.D., M.P.H., has evaluated the use of artificial intelligence among more than 15,000 family medicine residency trainees across the U.S. and found that in October 2025, 71% of residents had used AI for work in the past month. The paper is published in the journal Family Medicine.

“We knew trainees were using AI, but we didn’t know how many and what the most common uses were until now,” says Barr, associate professor of medicine (general medicine) and founding program director of the Yale New Haven Health Family Medicine Residency Program. “This was a great opportunity to really understand the AI phenomenon in residency training.”

There are a variety of AI tools with applications in the practice of medicine, ranging from tools that help transcribe patient-clinician conversations to databases for searching medical references to programs designed to support clinical decision-making.

The researchers also surveyed the residents about whether they received any training or continuing medical education on the tools. Of the 71% who reported using AI, only 36% said they had received training in the last six months.

A standard component of medical school and residency training is learning to critically appraise medical literature and research. Relying on AI with little to no training can lead to negative outcomes for trainees, such as accepting the AI tool’s output as correct without applying critical thinking, losing already learned skills and never learning the skill.

“Part of training is learning how to use information wisely and efficiently and ensuring you’re looking things up appropriately to get the right information,” says Barr. “We’re still learning about how AI affects the way we think about clinical decision-making, which is key for diagnosis and treatment.”

While AI tools can be useful, proper training is needed to help residents assess and evaluate the information, says Barr. She notes how medical knowledge has expanded rapidly and continues to accelerate, and how crucial it is for trainees to have the skill set to evaluate information for each patient and situation.

“If trainees use AI tools, we need to teach them how to utilize the tools correctly and in a way that benefits patients,” she urges. “You don’t pick up a stethoscope and use it without being shown how to use one, and this is no different.”

The solution is to incorporate AI training with critical appraisal in medical school and residency training curricula, says Barr. “The question for institutions is: When is it appropriate to incorporate it in the medical education training process?”

Barr will further research the use of AI using the 2025 and 2026 survey results, focusing on outcomes related to whether trainees use AI and how they use the technology.

Who’s behind this story?

Lisa Lock

BA art history, MA material culture. Former museum editor, paramedic, and transplant coordinator. Editing for Science X since 2021.

Full profile →

Andrew Zinin

Master’s in physics with research experience. Long-time science news enthusiast. Plays key role in Science X’s editorial success.

Full profile →