<p>Manufacturing systems, due to their complexity and size, can often generate large quantities of data. The organization of this data for the purposes of sustainability assessment and improvement is a difficult and laborious endeavor. Sustainability data covers all facets of production, requires significant amounts of processing before it can be used, and is acquired from highly fragmented sources. There is a clear need for structured methods to develop an integrated knowledge base that can present and analyze this data effectively and inform decision support. To this effect, domain-specific large language models (DS-LLMs) are proposed as a viable solution. When sufficiently grounded in well-curated knowledge, LLMs can offer powerful capabilities for sustainability reasoning. This paper presents a framework for curating and prototyping knowledge bases suitable for LLM-assisted sustainability studies. The manufacturing sector being studied here is electroplating. Approximately 1000 curated sources in the domains of government documents, academic studies, and industrial reports were organized into seven major categories and over 100 subtopics. The resulting dataset integrates regulatory, scholarly, and industrial materials into a structured knowledge base suitable for reasoning and retrieval. Targeted data augmentation and gap analysis cycles were implemented to ensure comprehensive coverage, as iterative testing with NotebookLM provided refinement of dataset quality and evaluation of the framework’s analytical performance. NotebookLM served as an AI-assisted environment for summarization, reasoning, and gap identification, transforming raw collections of data into a coherent, sustainability-aligned knowledge system to support sustainability analysis. A case study illustrates that the curated knowledge base enables systematic sustainability assessment and identifies improvement strategies for electroplating operations through examination with NotebookLM. This research contributes a systematic framework for preparing, organizing, and curating domain-specific manufacturing knowledge to support AI-assisted sustainability analysis. The use of electroplating is intended to illustrate the applicability of the framework, and the systematic and methodological design enables replication for the development of similar knowledge bases for other manufacturing sectors.</p> Graphical abstract <p></p>