Elena Iliadis
Washington, D.C., United States

The better part of medical school, I would argue, is learning to speak the language.

The language of medicine is nuanced: there is language of diagnosis and of prescribing, language between medical professionals, and importantly, language within the physician-patient relationship. There are the foundational languages, like physiology and immunology, and the experience-based languages, like clinical reasoning and morning rounds. As I have moved through medical school, some languages have felt foreign, some required daily practice to strengthen the muscle, and some came so naturally I hardly noticed my own fluency. Yet to truly speak the language of medicine, you must synthesize these different dialects into one cohesive voice.

Like any living language, the language of medicine evolves with cultural shifts, scientific advances, and, most recently, the emergence of artificial intelligence (AI). As AI becomes increasingly woven into medical education, we must integrate it in a way that strengthens rather than substitutes for our humanity.

ChatGPT gained momentum during my first year in medical school, making my class one of the first to have generative AI at our fingertips. Admittedly, I have always lagged behind my peers in adopting new technologies, and at first, it seemed more of a luxury than a necessity. The first time a classmate suggested we “ask AI” during a weekly small group session, I remember joining the playful jokes. It felt inherently unserious to bring a chatbot into our discussion of arrhythmias, and no one seemed eager to trust its read of an electrocardiogram over our own. It also had a reputation for inaccuracies and oversimplifications, and nuanced clinical reasoning felt beyond its scope.

Yet over my first year, many of my peers began embracing generative AI as a standard resource—handing it problem sets when stumped, asking it to outline writing assignments, or using it to generate a differential diagnosis. Starting your sentence with “ChatGPT said” became common, as if synonymous with other resources like UpToDate or PubMed. Seemingly overnight, a silent consensus formed that the flaws of AI were outweighed by its efficiency. AI was no longer just a tool; it was a new language entering medical education.

During my preclinical years, AI was never formally acknowledged as a resource or integrated into our curriculum, but it was never discouraged or regulated either. And despite my own uncertainties, I could not ignore its usefulness. It could simplify complex subject matter, answer follow-up questions like a personal tutor, and communicate in your own preferred learning style. In medical school, when the sheer volume of information could feel insurmountable, ChatGPT could translate medicine into a digestible language. It was easy, available, and customizable—so it was no mystery why the language of AI spoke to medical students.

When I began clinical rotations, I assumed AI’s role in my training had passed its peak. Preclinical knowledge might be translatable, but clinical medicine felt fundamentally human. That assumption lasted days into my first rotation, when I encountered ChatGPT’s more clinically focused counterpart: OpenEvidence. Marketed as an “AI copilot for doctors,” OpenEvidence offers a HIPAA-compliant platform to ask point-of-care questions. Backed by clinical organizations and respected journals, it is positioned as a trusted, validated resource for physicians.

Unlike most generative AI, OpenEvidence speaks like a fellow clinician, assuming a foundation of medical knowledge in its audience. Where standard AI reflexively simplifies, OpenEvidence synthesizes. It does not translate medicine into simpler terms; it participates in the conversation of clinical reasoning. Simply put, it speaks the language of medicine, and it assumes you speak it too.

As a third-year student, OpenEvidence proved invaluable during rotations. I used it to outline surgical procedures, apply screening guidelines to individual patients, and understand criteria for anticoagulation or repeat imaging. As each rotation introduced unfamiliar territory, OpenEvidence offered a consistent dialogue with the literature, a key asset in my pursuit of clinical fluency.

Still, AI’s growing presence in medical education comes with clear limitations. During rotations, I often heard praise for tools like Doximity or Suki AI, which can listen to patient encounters and automate documentation. But I also heard concerns: that these systems record every detail indiscriminately, struggle to distinguish relevant history from casual conversation, and raise serious questions about privacy and patient trust. These warnings reinforced that while AI has the potential to make medicine more efficient, we must remain critical of its flaws and cautious of its proximity to our patients.

Rotations also highlighted how much of medicine remains beyond AI’s reach. Of course, AI could not help me remove staples, or listen for crackles in a patient’s lungs. But most importantly, it could not offer the presence I could, whether comforting a tearful patient or navigating a difficult goals-of-care conversation. It could not counsel or empathize, de-escalate tensions or set boundaries. While there were many things I could not yet do as a medical student, one thing I could always offer was the language of humanity. And that language is irreplaceable.

As I enter my final year of medical school, I have made peace with the permanence of AI in my training. The challenge for medical education now is teaching my generation of physicians how to incorporate AI in a way that enhances but does not replace evidence-based medicine.

Across medical school and residency, students and educators must learn to use AI intentionally and professionally. Tools like ChatGPT can streamline and personalize learning, but if used uncritically, they can just as easily undermine it. OpenEvidence can scan vast bodies of literature and synthesize evidence, but it lacks situational discernment.

AI may improve efficiency, but it cannot replace the human capacities for nuanced clinical reasoning and genuine compassion. It is only useful in explaining complex pathology if we use that understanding to better educate our patients. It is only helpful in streamlining documentation if we spend those saved minutes listening at the bedside.

AI is now part of the language of medicine. But as we train the next generation of physicians, we must use it to strengthen our clinical voice. It cannot speak for us.