Environmental Impact Assessment (EIA) is a fundamental instrument for integratingenvironmental considerations into decision-making. However, its effectiveness faces persistentchallenges, particularly regarding the Determination of Impact Significance, widely recognizedas the core of the EIA process, yet frequently lacking systematization and transparency inEnvironmental Impact Statements (EIS). This research investigated the potential of ArtificialIntelligence (AI), through Natural Language Processing (NLP) and Large Language Models(LLMs), to support the documentary analysis of how the Determination of Significance isaddressed in Brazilian EIS. The methodology involved a systematic literature review thatresulted in a set of six key assessment questions for evaluating significance practices in EIS;the empirical analysis of 35 Brazilian EIS using this instrument; the development of anexperimental system based on Retrieval-Augmented Generation (RAG) with two commercialLLMs (GPT-4o and Claude 3.5 Sonnet); and a comparative evaluation between model-generated responses and human reference analysis. The empirical analysis revealed significantgaps in Brazilian practice, with only 22 of the 35 EIS providing a definition of what theyconsider significant and 16 cases where it was not possible to verify the coherence between theadopted definition and the judgments made. The system demonstrated compatibility withhuman analysis for objective questions, with Claude 3.5 Sonnet showing better performance inidentifying formal definitions and conceptual structures. Limitations were identified, includingGPT-4o's tendency to infer beyond the explicit text and Claude's generation of inconclusivesummaries in some cases. It is concluded that NLP and LLMs can support the documentaryanalysis of the Determination of Significance in EIS, with potential for initial screening andinformation retrieval, without replacing specialized human analysis, which remainsindispensable when documents lack transparency in their judgment criteria.