Abstract: Medical device quality control is fundamental to ensuring reliable device performance and patient safety. Conventional QC record management relies heavily on manual data entry, leading to multiple persistent challenges, including low efficiency, insufficient standardization, delayed knowledge updates, and high costs for traceability and post-incident investigation. With rapid advances in large language models (LLMs) in natural language processing, long-document understanding, and domain knowledge retrieval, a new human–AI collaboration paradigm has emerged for building and optimizing end-to-end QC systems across the medical device lifecycle, and they are well suited for exploratory applications under a workflow combining controllable automated generation with manual review and verification.. Focusing on the medical device QC workflow, this paper systematically reviews the overall architecture, key techniques, and core application scenarios of LLM-based QC systems for medical devices, and summarizes system implementation pathways and key engineering considerations. In addition, grounded in existing safety frameworks, we further discuss challenges and risks related to data security, factual reliability, interpretability, and ethical and regulatory compliance., and discusses potential application scenarios and implementation conditions in areas such as predictive maintenance and digital twins, aiming to provide references for both academic research and industrial deployment in this area.