Finding the right clinical trial for a patient can be like finding a needle in a haystack. Doctors spend countless hours sifting through complex medical data and trial criteria, a process that's both time-consuming and prone to errors. But what if AI could lend a hand? New research explores how large language models (LLMs) can be used to streamline this crucial process. Researchers tested six different LLMs, focusing on smaller, open-source models that can run locally in hospitals, preserving patient privacy. These AI assistants were tasked with generating search queries based on synthetic patient data, which were then used to retrieve relevant clinical trials. Surprisingly, the smaller LLMs performed as well as, or even better than, queries crafted by medical experts, especially when combined with the original patient notes. Furthermore, these AI-powered searches were lightning fast, returning results in mere seconds. This opens exciting possibilities for faster, more efficient clinical trial matching, ultimately connecting patients with potentially life-saving treatments. While larger models like GPT-4 were also tested, the smaller, open-source models offer a compelling balance of performance, efficiency, and data security, potentially paving the way for wider adoption in real-world healthcare settings. This research demonstrates the promise of AI not to replace doctors, but to empower them with powerful tools to improve patient care.