Abstract: Large language models（LLMs） have achieved groundbreaking advancements in artificial intelligence and are poised to revolutionize military applications. Traditional military force resource recommendation systems rely on target IDs, force resource IDs, and historical interaction data for predictions, but suffer from poor data utilization, weak transferability, and cold-start problems. To enhance the capability and efficiency of military force recommendation systems and improve the scientific rigor of target strike planning decisions, this study analyzes LLMs applications through the lens of the military recommendation paradigm. We examine LLM’s potential across feature engineering, feature encoding, and scoring prediction stages, demonstrating its evolution from an auxiliary tool to a comprehensive
replacement. Comparative analysis of different application scenarios reveals distinct advantages. Finally, we summarize both opportunities and challenges for LLMs implementation in military recommendation systems. This research anticipates LLMs enabling more intelligent military decision-making methods while compensating for current technological and force structure limitations through AI-driven advantages.