Imagine a doctor prescribing medication not just based on test results, but also by deeply understanding your medical history. That's the promise of new research using Large Language Models (LLMs) to enhance medication recommendations. Traditionally, AI models for this task relied heavily on structured data like medical codes. Think of these codes as shorthand for diagnoses, procedures, and medications. While useful, this approach often overlooked the wealth of information contained within unstructured data – the detailed notes doctors write about each patient. This new research taps into that unused potential. Researchers found that LLMs, known for their powerful language processing abilities, can extract valuable insights from lengthy clinical notes, even deciphering complex medical terminology. By combining these insights with traditional medical codes, they created a more holistic view of the patient's condition. The results are impressive. In experiments using two real-world datasets (MIMIC-III and data from Chia-Yi Christian Hospital), the LLM-enhanced models showed a significant improvement in medication recommendation accuracy. One model, G-BERT, performed noticeably better when combining text and code insights. Others, like COGNet and GSVEMed, demonstrated a comparable ability to recommend medications using only text, rivaling the performance of code-based methods. This research suggests a future where AI can more accurately tailor medication recommendations to individual patient needs by understanding both the “what” (codes) and the “why” (clinical notes) of their medical history. However, challenges remain. Processing long clinical notes efficiently and integrating text insights seamlessly into existing models require further innovation. Moreover, ensuring patient safety by minimizing potential drug-drug interactions is crucial. As researchers address these challenges, we can expect LLMs to play an increasingly important role in personalized medicine, leading to safer and more effective treatments.