Metaphors are powerful tools in discourse, shaping how we interpret events, issues, or concepts. Consequently, developing methods for automated and comprehensive analysis of metaphors in large-scale textual data becomes a critical task. In this context, the ability of artificial intelligence to generate, interpret and relate metaphors to the cognitive-pragmatic context of speech becomes a particularly relevant question. Modern AI tools are predominantly developed by Western research teams. They rely on foreign technologies and are based on the methodology of foreign cognitive linguistics. They are also developed using English-language corpora. As a result, they depend on foreign technology stacks, they are methodologically rooted in Western cognitive linguistics and trained almost exclusively on English-language corpora. This study therefore aims to provide a solution for identifying and analyzing metaphors in Russian-language texts by developing a novel tool methodologically grounded in Russian linguistics and built using domestic technological resources. The approach draws on the theory of metaphorical speech impact. Within this theory, the functional potential of any metaphor in discourse unfolds through four functions (representational, evaluative, persuasive, and suggestive) across cognitive, semantic, and communicative levels and is subsequently quantified through specific metaphoricity indices. Technically, the solution employs a hybrid approach, combining prompt engineering, rule-based code, and access to the YandexGPT generative model through its cloud API within a Python 3.10+ environment. The methodological procedure comprises three sequential stages from model training to metaphor analysis. In the first stage, metaphor detection and modeling, the model learns to identify metaphors, classify them by source and target domains, and construct “A is B” metaphorical models. Building on this, the second stage, metaphor classification and assessment, involves identifying the specific metaphor type (ontological, orientational, structural), determining its intensity (conventional, moderate, novel), and assessing its evaluation (negative, positive, neutral). The third and final stage is quantitative analysis and interpretation. This involves calculating the core indices of the impact of metaphorical speech – density, intensity and typology indices – and providing a comprehensive interpretation of the results. This research culminates in the development of the Metaphor Analytics Software, capable of automated metaphor detection, classification, and analysis. This program effectively fills a significant gap in the system of NLP tools for analyzing Russian-language metaphors.