Medical students have a new study partner. It's available at 2 a.m., answers follow-up questions without hesitation, and never runs out of patience. The problem is that it can also get things wrong and sound confident when it shouldn’t.

A recent survey of medical students explored how they are using AI to study, from revision and concept explanation to clinical reasoning, prompting and verification. The research, conducted by McGraw Hill, called What medical students' AI study habits mean for teaching, assessment and clinical readiness, found that 73% of students surveyed use AI daily, while 91% said it has had a positive impact on their studies.

AI now sits alongside lecture notes, textbooks and question banks as a routine way to explain concepts, revise and test understanding.

For faculty, the question is no longer whether students will use AI, but what happens to their thinking when they do. Medical education depends on students learning to question information, connect knowledge, develop reasoning skills, and make sound clinical judgements.

The opportunity for educators is to help students develop habits that allow them to use AI while keeping critical thinking and reasoning visible. A useful starting point is the learning objective itself: what do students need to learn, and then what role, if any, can AI play in helping them to get there? Here are five practical ways faculty can respond to the ways students are using AI:

1. Find out how, when, and where students are already using AI in their learning

It’s difficult to guide a behaviour that can’t be seen. The research found that 74% of students said they use AI to explain difficult concepts and revise, while 61% use it conversationally to ask follow-up questions.

That might happen before a lecture, while working through a difficult topic, or when a faculty member isn’t around. Rather than trying to monitor every interaction, faculty can create opportunities for students to talk about how they are using AI: what they used to prepare for a session, where they find it useful, and where its response left them uncertain.

These conversations can give faculty a clearer view of students' learning processes and create a natural opening for discussing good practice.

2. Make expectations clear

As AI becomes part of students’ everyday study routines, expectations around when and how to use it vary between courses, assignments, and institutions.

Medical schools are at differing stages of AI adoption, policymaking, and guidance. Some schools, such as the University of Virginia (UVA) School of Medicine, have students practice using AI to diagnose and develop treatment plans for fictitious patients [1], while others are defining permitted uses and expectations around citation, and others are still developing their approach.

While individual faculty do not control institutional policy, they can make expectations clear within their own teaching. For a particular activity or assignment, for example, faculty can explain where AI can be used, what students need to do themselves and whether its use should be acknowledged.

3. Build verification and source evaluation into learning

Students are already aware that AI can and does get things wrong. But that awareness doesn't always translate into consistent verification: 88% of those surveyed said they had received incorrect or misleading AI responses, yet only six in ten said they consistently verify AI outputs.

One practical response is to make verification and source evaluation part of the learning activity itself. For example, if the learning goal is to evaluate information and evidence critically, an AI-generated differential diagnosis, even a poor AI-generated one, can give students something to test, question, and verify. Give students a checklist for them to evaluate it:

Check the response

  • What is accurate?
  • What is missing?
  • What would you challenge?

Check the evidence

  • What evidence would you use to verify it?
  • Is that evidence current and traceable?
  • Has it been reviewed by someone with the right expertise?

Used this way, the AI output becomes the beginning of the thinking rather than the end of it.

4. Keep students’ reasoning visible during learning

Finding an answer and reasoning through a clinical problem are not the same thing. The research suggests students are already using AI for activities to aid clinical thinking: 77% of students surveyed said they simulate clinical reasoning through AI, including working through differential diagnoses.

Faculty can create opportunities for students to show their own thinking before bringing AI into the activity. During a case discussion, for example, students might commit to an initial diagnosis or clinical decision and explain the reasoning behind it before comparing their thinking with an AI-generated response. They can then consider what the AI has added, whether it changes their view and why.

The important question isn’t whether the student agrees with AI; it's whether they can evaluate the information and justify their own clinical reasoning.

5. Assess the process, not just the final answer

The same principle applies when students are being assessed. As AI becomes part of everyday study, it’s worth asking whether a task still reveals the knowledge, reasoning and judgement it was designed to assess.

A useful question is: Could a student produce a plausible answer with AI without demonstrating the capability this task is intended to assess?

If the answer is yes, the task might not need a full redesign. A small change could make the student’s process easier to evaluate. For example, where AI use is permitted, faculty could ask students to document the steps they followed to reach their answer: how they used AI, which outputs they questioned or rejected, what sources they consulted, and how they arrived at their final judgement.

In a case-based assessment, students could also be asked to justify why one diagnosis is more likely than another, explain what additional information would change their decision, or defend the next step they would take in the patient’s care.

The aim is to assess not only what students produce, but how they evaluate information, make decisions and reach their own clinical judgement.

Keep the thinking at the center

AI will continue to change, and so will the ways medical students use it. Faculty don’t need to understand every new tool or become AI specialists to respond constructively.

The opportunity is to strengthen the habits students will need regardless of the technology in front of them: checking evidence, questioning information, recognising uncertainty, explaining their reasoning and knowing when an answer deserves a second look. Those are the habits that will matter well beyond the classroom.

Want to explore more about how medical students are using AI and what it means for teaching, assessment and clinical readiness?

Read the full McGraw Hill report: What medical students’ AI study habits mean for teaching, assessment and clinical readiness by completing the form below.