Tram Huynh – HowMedWorks

Physician associate/assistant (PA) school has a reputation for being intense, and it's well-earned. Didactic year throws an overwhelming volume of information at you that’s compressed into 12 to 15 months. You're expected to absorb pharmacology, pathophysiology, clinical medicine, and diagnostic reasoning at breakneck speed while staying current on evidence-based guidelines. This article was written by Tram Huynh, a current PA-S1 at Long Island University. This is where artificial intelligence (AI) became a game-changer for me. Throughout my first semester, I used Google’s NotebookLM, Anthropic’s Claude, OpenAI’s ChatGPT, and OpenEvidence to transform dense lectures and clinical guidelines into study materials that actually matched how I learned. Keep reading to find out what worked, what didn’t, and how I’d use each tool if I had to start didactic year all over again. Why PA Students Should Care About AI You don’t need AI to succeed in PA school. Many people still thrive with highlighters, whiteboards, and motivation fueled by iced coffee. However, AI study tools can help significantly cut down on studying time, which matters when your schedule is nonstop. More importantly, AI will continue to play a growing role in healthcare, as it's becoming increasingly integrated into every aspect of medicine: Radiologists currently use AI to flag potential abnormalities in imaging studies. Electronic health records now incorporate algorithms to identify patients at risk for sepsis or readmission. Clinical decision support tools powered by machine learning help providers consider diagnoses they might have missed. 1. Google NotebookLM: The Audio-First Study Buddy What It Does Well Google's NotebookLM has one feature that absolutely shines: audio summaries. If you’re an auditory learner, you’re in luck. After uploading lecture slides, notes, or PDFs, you can generate an "Audio Overview," which is a podcast-style discussion in plain, friendly language. If you have a commute, or if your eyes stop functioning after eight hours of lectures, this is a genuine lifesaver. It turns your study materials into something you can review while walking to class, cooking, or pretending to go to the gym. It also provides exact source citations, which makes checking accuracy easy. Limitations NotebookLM has strict file size limits: PDFs are capped at 500 KB and word count caps across all sources. This is problematic when a single pharmacology lecture alone can exceed these limits. Additionally, NotebookLM doesn’t produce strong clinical vignette questions. The questions are simple and fact-based, not the multi-step reasoning style ones that you’ll see on school exams and the PA National Certifying Examination (PANCE). Best Use Google NotebookLM is best used as a supplementary tool (i.e. for audio reviews and quick clarifications, not as a main study engine). 2. Anthropic Claude: Heavy Lifter for Big Documents What It Does Well Anthropic’s Claude.ai can process large volumes of information. It can handle entire textbook chapters, comprehensive lecture slide decks, and multiple documents simultaneously. Its "Projects" feature allows you to upload all materials for a specific course to make comprehensive practice exams that integrate information across multiple lectures. Practice Exam Quality I’ve found that Claude is very good at generating clinical vignette-style questions that closely mirror PANCE format. With good prompting, it provides questions with realistic patient presentations (e.g. relevant history, physical exam findings, and laboratory values) that require multi-step reasoning in order to figure out the correct answer. Pro tip: When it comes to prompting any AI models, you want to be specific and give clear instructions on what types of questions to create. For instance, try uploading a few high-quality example questions and ask Claude.ai to match that style and difficulty level. The more constraints you provide (e.g. question format, difficulty level, specific topics, number of distractors [i.e. incorrect answer choices that require true understanding to eliminate]), the more tailored and useful your practice questions will be. So instead of a basic question like "What is the first-line treatment for hypertension?", Claude can generate: "A 58-year-old African American male with a BMI of 32 presents for follow-up. His blood pressure readings over three visits average 156/94 mmHg. He has no history of diabetes, chronic kidney disease, or cardiovascular disease. Laboratory studies show normal creatinine and potassium. Which antihypertensive medication is most appropriate for initial therapy?" Creating Anki Decks Claude is also great at generating large batches of flashcards in Anki- and Quizlet-compatible formats, including cloze deletions (i.e. hiding one or more words or fill-in-the-blank). If you rely heavily on spaced repetition, this can save you hours! Limitations Extremely lengthy uploads may still be difficult; entire textbooks may need to be broken into sections. Also, like with all AI tools, Claude's medical knowledge has a cutoff date so make sure to verify the most current clinical guidelines or recently approved medications. Best Use Claude.ai excels at processing large amounts of source material and is best used for building exam banks and generating flashcards efficiently. 3. OpenAI ChatGPT: The Versatile Exam Generator What It Does Well ChatGPT is an AI chatbot that produces high-quality clinical vignette questions that align well with those on the PANCE. The questions typically include appropriate patient demographics, relevant clinical details, and plausible distractors. Beyond questions, ChatGPT can: Generate Anki flashcards; Create summary tables; Develop mnemonics; and Create images if you’re a visual learner! If sodium-potassium channels are frying your brain, ChatGPT can rephrase the explanation with analogies, diagrams, or simpler language until it clicks! OpenAI’s ChatGPT, particularly the GPT-4 model, has consistently been my go-to for exam prep. I can ask it for 50 practice questions on one topic and get solid vignettes in seconds. “Custom Instructions,” makes it even more powerful, as you can set it to always generate PANCE-style questions with explanations for correct and incorrect answers. It’s ridiculously efficient. Limitations With ChatGPT Plus, you can upload documents, though with some limitations. The file handling isn't as robust as Claude's, very large documents may need to be split, and the context window (how much information the AI can actively work with at once) is more limited. However, for individual lecture slide decks or notes from a single class session, ChatGPT works great. One perk is that ChatGPT's integration with plugins and web browsing (depending on your subscription), can help verify current clinical guidelines or look up recently updated information. However, it isn’t perfect and sometimes does include subtle inaccuracies, so always double-check anything that feels off. Best Use ChatGPT is best for producing high-quality practice questions and breaking down confusing concepts. 4. OpenEvidence: The Evidence-Based Medicine Powerhouse What It Does Well OpenEvidence is specifically designed for healthcare professionals and students who need quick access to evidence-based medical information. Unlike general AI chatbots, OpenEvidence searches through peer-reviewed medical literature, clinical guidelines, and systematic reviews to provide answers backed by current research. The platform excels at: Providing evidence-based answers with direct citations. Suggesting differential diagnoses. Comparing treatment modalities. Staying current with medical guidelines. Identifying key primary studies and landmark trials relevant to a clinical question. Highlighting areas of consensus and controversy within the literature. OpenEvidence is particularly useful for PA students because it provides quick, evidence-based answers with clear citations. Additionally, you don't have to worry as much as you would with a general online search, for instance, because the responses are supported by cited, peer-reviewed evidence. Pro tip: For my capstone project (aka the big research paper required to graduate from PA school), I used OpenEvidence to help identify key articles that I might have missed that were relevant to my topic. —Phoebe Limitations OpenEvidence is highly specialized. It's focused on clinical medicine and evidence-based practice so it may not the best for study tasks like creating flashcards or practice questions. Additionally, while the citations are extremely helpful, you still need to make sure to evaluate the evidence presented. Not all studies are created equal, and OpenEvidence's coverage is not necessarily exhaustive, despite drawing from a wide range of peer-reviewed medical literature. Best Use OpenEvidence is best for when you need evidence-based answers to clinical questions, want to verify treatment guidelines, or need to cite medical literature in assignments and presentations. Comparative Analysis: Which Tool for Which Purpose? Use Case Best Tool Large documents Claude Practice exams Claude and ChatGPT (tie) Audio learning NotebookLM Anki decks Claude and ChatGPT (tie) Explaining concepts ChatGPT and OpenEvidence (tie) Evidence-based answers OpenEvidence Current clinical guidelines OpenEvidence Practical Workflow Recommendations The most effective approach isn't choosing one platform but strategically using multiple tools: Claude for building the base. Upload full lecture sets into Projects → generate long practice exams → create bulk Anki cards. ChatGPT for weak spots. Identify gaps → generate targeted questions → ask for explanations in different formats. NotebookLM for commute review. Upload summaries → generate audio overviews → listen while doing anything other than sitting at your desk. OpenEvidence for clinical clarity. Look up treatment guidelines → verify evidence for case presentations → understand the reasoning behind clinical decisions. Final Thoughts In the coming years, AI will play an even larger role in differential diagnosis generation, treatment planning, medication management, and patient education. Medical training is also shifting toward AI-driven simulations and virtual patient encounters. Knowing how to prompt, interpret, and verify these tools is quickly becoming a core skill of clinical practice. Students who learn to leverage AI now are simply preparing for the realities of future practice. Food for thought: AI won’t replace providers, but providers who use AI will replace those who don’t. AI can’t replace the human parts of medicine: listening, empathy, physical exams, and trust. However, clinicians who can integrate AI thoughtfully will undoubtedly provide more efficient, informed care. The evolution of AI in medical education isn't something happening to PA students, it’s something the current generation of students will help shape. By thoughtfully experimenting with these tools now, providing feedback to developers, and thinking critically about effective implementation, today's students are helping share the future of medical training and education. Conclusion To be completely honest, the only way I have time to write this article is with a lot of time management and a true appreciation for AI. It helped me organize, summarize, and outline my study as well as article materials. You don’t have to use AI, but if you want to lighten your workload and prepare for the future of healthcare, it’s worth experimenting with. AI isn’t here to replace you. It’s here to support you while you become the compassionate, human-centered provider your patients will depend on. Use the tools, experiment with them, and build a system that supports your learning style while preparing for an increasingly tech-integrated healthcare system. Hopefully, you found this guide helpful! References U.S. Food and Drug Administration. Artificial intelligence and machine learning (AI/ML)-enabled medical devices. Accessed [Month Day, Year]. https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices Hsu A. AI is helping radiologists read images faster and more accurately. NPR. Published 2023. Accessed [Month Day, Year]. https://www.npr.org/ Roberts S. A.I. may help doctors spot cancer but comes with its own risks. New York Times. Published 2023. Accessed [Month Day, Year]. https://www.nytimes.com/ Palmer P. Hospitals are increasingly using AI to help flag sepsis and other risks. STAT News. Published 2022. Accessed [Month Day, Year]. https://www.statnews.com/ Harvard Health Publishing. How artificial intelligence will change medical practice. Harvard Health. Published n.d. Accessed [Month Day, Year]. https://www.health.harvard.edu/ American Medical Association. How AI is reshaping clinical workflows. AMA. Published 2023. Accessed [Month Day, Year]. https://www.ama-assn.org/ World Health Organization. Ethics and governance of artificial intelligence for health. World Health Organization; n.d. Accessed [Month Day, Year]. https://www.who.int/