A research team led by Professor Lin Haotian from the Zhongshan Ophthalmic Center at Sun Yat-sen University has developed a novel research paradigm centered on large language models (LLMs) as intelligent assistants in medical AI innovation. The study, recently published in Cell Reports Medicine, presents the first randomized controlled trial validating the effectiveness of LLMs in empowering clinicians to overcome technical barriers in AI-driven medical research. As artificial intelligence increasingly integrates into healthcare, clinical researchers face significant challenges, including high technical barriers and limited cross-disciplinary expertise in AI. To address these obstacles, the team conducted a rigorous randomized controlled trial to systematically evaluate how LLMs assist physicians without prior AI experience in designing and executing medical AI research projects. The results demonstrated that LLMs significantly enhance the feasibility of research proposals and reduce the time required to complete projects. Doctors using LLMs were able to independently develop robust AI research plans, achieving outcomes comparable to those of experienced AI researchers. The study also revealed a positive "skill transfer" effect, where physicians gradually improved their understanding of AI concepts and methodologies through interaction with the models. However, the trial also identified potential risks, particularly over-reliance on LLM-generated content, which could compromise the accuracy and originality of research. These findings underscore the need for responsible and structured use of LLMs in medical research. In response, the team developed a practical, evidence-based framework called the "CPGI" prompt construction guide—standing for Context, Purpose, Guidance, and Iteration. This structured approach provides clinicians with a clear, logical method to interact with LLMs, promoting safe, effective, and reproducible use in research settings. The study marks a significant step toward democratizing medical AI innovation, enabling non-technical clinicians to participate meaningfully in the development of next-generation AI tools for healthcare. The findings offer a scientific foundation for future guidelines on integrating LLMs into clinical research workflows.