For years, medical schools were teaching AI like someone poking at a single trumpet and calling it jazz. This new paper tries something more ambitious: it assembles the whole orchestra and asks, in plain terms, what a future doctor actually needs to know before an algorithm starts whispering into the exam room.

That paper, a 2026 scoping review in npj Digital Medicine, looked across 4,071 records and pulled in 54 studies from 22 countries to build a competency map for undergraduate medical education [1]. The result was not a cute little checklist. It was a seven-domain taxonomy covering AI ethics, law and regulation, professionalism, clinical applications, critical appraisal of AI output, research and innovation, and theory and foundations. In total: 37 competencies and 170 learning objectives. If you were hoping for "just teach them ChatGPT and move on," the numbers tell a different story.

Follow the Paper Trail, Not the Hype

According to the authors, the strongest themes across the literature were not coding tricks or startup fever dreams. They were ethics, legal oversight, and the ability to critically appraise AI output [1]. That matters. A lot.

Because the sales pitch around medical AI usually sounds like this: smarter diagnosis, faster workflows, fewer mistakes, maybe a robot sidekick that never sleeps and does not eat stale granola bars in the call room. But when pressed, the real question is less "Can doctors use AI?" and more "Can doctors tell when AI is wrong, biased, overconfident, or being weird in a statistically sophisticated way?"

That is not a minor detail. That is the whole job.

The review also points out an awkward fact: much of the competency literature is still editorial or opinion-based rather than built from hard validation studies [1]. In other words, a lot of people have strong thoughts about what medical students should learn, but the evidence for how best to teach it is still catching up. AI education, at least for now, has some big-group-chat energy.

The New Bedside Manner Includes Skepticism

To understand why this matters, zoom out. AI in healthcare already touches diagnostics, treatment planning, drug development, and patient monitoring [6]. Clinical decision support systems are designed to help clinicians make decisions at the point of care, not replace them outright [7]. The distinction matters because medicine is full of edge cases, missing context, and humans who inconveniently refuse to behave like benchmark datasets.

Recent evidence only sharpens that tension. Stanford researchers reported in February 2025 that a chatbot outperformed physicians who had access only to internet search and references on nuanced clinical management questions, while physicians paired with the chatbot performed as well as the chatbot itself [5]. That is not a "doctors are finished" story. It is a "human plus machine might beat either alone, if the human knows what they are doing" story.

Which brings us back to the review. It is essentially arguing that AI literacy in medicine should look less like a coding bootcamp and more like safety training for a very clever, very confident coworker who occasionally improvises facts.

Everyone Wants AI in the Curriculum. The Curriculum Would Like a Word.

This paper lands in the middle of a broader scramble. The Association of American Medical Colleges says the share of MD- and DO-granting schools in the U.S. and Canada incorporating AI into their curricula rose from 53% in 2023 to 77% in 2024, and it is developing a national competency framework with a target release in fall 2026 [4].

Meanwhile, schools are already experimenting. AAMC reported in February 2025 that medical programs are using AI to generate quizzes, simulate patients, spot student struggles, and help draft evaluations, though with plenty of caution about oversight [3]. That makes sense. Medical education has never exactly suffered from too little complexity. Adding generative AI is like deciding your already chaotic kitchen also needs a flamethrower.

Other recent reviews echo the same theme. A 2024 BEME guide mapped AI's growing role in medical education and called for more coherent integration [2]. A 2025 updated scoping review in BMC Medical Education found the literature exploding after large language models went mainstream, while also stressing the need for faculty development and better evidence on outcomes [8].

So yes, AI is showing up in class. But the Hunt and colleagues review asks the more adult question: what should students actually be competent in before this stuff becomes routine?

The Useful Part Nobody Should Skip

The answer, according to this paper, is not "turn every medical student into a machine learning engineer" [1]. It is more grounded than that.

Students need enough foundation to understand what these systems are doing, enough ethical and legal awareness to use them responsibly, and enough clinical judgment to challenge the output instead of bowing to it like it descended from a glowing cloud. If you wanted to sketch that seven-domain sprawl for a curriculum committee without losing your will to live, a visual mapping tool like mapb2.io would honestly be more useful than yet another unloved spreadsheet.

The real contribution here is structure. Not certainty. Structure.

And that is why the paper is worth reading. It does not pretend the field is settled. It shows where consensus is forming, where the evidence is thin, and where medical schools may be teaching tomorrow's doctors to trust tools they do not yet know how to interrogate. In medicine, that is not a small curriculum gap. That is the plot.

References

1. Hunt VM, Sprehe LK, de Lomba WC, et al. What the AI era doctor should know: a scoping review of proposed artificial intelligence competencies for medical education. npj Digital Medicine. 2026. DOI: 10.1038/s41746-026-02761-9. Nature page: https://www.nature.com/articles/s41746-026-02761-9

2. Gordon M, Daniel M, Ajiboye A, et al. A scoping review of artificial intelligence in medical education: BEME Guide No. 84. Medical Teacher. 2024. DOI: 10.1080/0142159X.2024.2314198

3. Boyle P. AI in medical education: 5 ways schools are employing new tools. AAMC News. February 27, 2025. https://www.aamc.org/news/ai-medical-education-5-ways-schools-are-employing-new-tools

4. AAMC. Artificial Intelligence Competencies Across the Learning Continuum. Accessed May 23, 2026. https://www.aamc.org/about-us/medical-education/ai-competencies

5. Armitage H. Physician’s medical decisions benefit from AI, Stanford Medicine-led research finds. Stanford Medicine News Center. February 5, 2025; updated April 24, 2026. https://med.stanford.edu/news/all-news/2025/02/physician-decision-chatbot.html

6. Artificial intelligence in healthcare. Wikipedia. Accessed May 23, 2026. https://en.wikipedia.org/wiki/Artificial_intelligence_in_healthcare

7. Clinical decision support system. Wikipedia. Accessed May 23, 2026. https://en.wikipedia.org/wiki/Clinical_decision_support_system

8. Simoni AH, et al. Artificial intelligence in undergraduate medical education: an updated scoping review. BMC Medical Education. 2025. DOI: 10.1186/s12909-025-08188-2

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.