The role of using ChatGPT AI in writing medical scientific articles - Two years after

Two years after the first evaluation of ChatGPT (GPT-3.5) for medical scientific writing, this short communication reassesses its capabilities using the GPT-5 model. Under the same experimental protocol as in 2023, GPT-5 generated a coherent and factually accurate article, integrating verified references and a structured IMRaD format, with human oversight but without post-generation textual rewriting. The model demonstrated significant progress in factual reliability, reasoning, and technical architecture, reflecting the shift from static language prediction to retrieval-augmented reasoning. However, despite these advances, large language models remain dependent on human expertise for validation, ethical interpretation, and scientific judgment. The results highlight the potential of AI as a valuable assistant in medical writing, while reaffirming that authorship, responsibility, and critical analysis must remain human.