This paper presents a novel computer-aided synthesis method that mimics expert chemical reasoning by leveraging large-scale language models (LLMs). Instead of using LLMs to directly manipulate chemical structures, we leverage their ability to evaluate chemical strategies and guide search algorithms toward chemically meaningful solutions. We demonstrate this paradigm through two fundamental tasks: strategy-aware retrosynthetic planning and mechanism elucidation. In retrosynthetic planning, we specify a desired synthetic strategy in natural language (ranging from protecting group strategies to overall feasibility assessments) and use traditional or LLM-guided Monte Carlo tree searches to find pathways that satisfy these constraints. In mechanism elucidation, LLMs combine chemical principles with systematic exploration to guide the search for plausible reaction mechanisms. This method is robust across a wide range of chemical tasks, and recent larger models demonstrate increasingly sophisticated chemical reasoning. This work presents a new paradigm for computer-aided chemistry that combines the strategic understanding of LLMs with the precision of traditional chemical tools, opening the possibility of more intuitive and powerful automated chemical systems.
