This research investigates how ESG-related narratives influence the volatility of financial indices, leveraging text mining and machine learning techniques to build predictive models. While ESG factors have become increasingly important in investment decision-making, the relationship between qualitative narratives—drawn from reports, news, and social media—and ESG index volatility remains underexplored. The study integrates unstructured textual data with financial time series to capture how ESG sentiment, reputational risks, and emerging sustainability issues impact market behavior. Building on natural language processing and machine learning algorithms (such as LSTM, XGBoost, and Random Forest), the model analyzes sentiment patterns, topic detection, and narrative trends to predict fluctuations in ESG indices. The main objective is to enhance financial risk management by offering an innovative framework that connects narrative analysis with volatility forecasting. This approach addresses key challenges such as linguistic variability and the subjective nature of ESG narratives, providing more accurate signals for market participants. The research offers practical contributions by enabling investors, financial institutions, and policymakers to anticipate market movements influenced by ESG-related events. Moreover, it supports the development of sustainable financial markets by improving the ability to monitor environmental, social, and governance risks. This framework is especially relevant for emerging markets, including Latin America, where sustainable finance is rapidly gaining importance but still faces challenges related to transparency and data availability.