This paper introduces SARF (Sentiment-Augmented Random Forest), a novel approach for stock market prediction that integrates sentiment analysis from FinGPT, a generative AI model, with the traditional Random Forest algorithm. The authors argue that relying solely on empirical analysis yields unsustainable results, and machine learning, particularly Random Forest, can enhance predictions. SARF leverages FinGPT to extract nuanced financial sentiments from textual data, which are then incorporated as features into the Random Forest framework. The methodology is evaluated against conventional Random Forest and LSTM models, demonstrating that SARF achieves an average accuracy improvement of 9.23% and lower prediction errors in forecasting stock market movements. The study highlights the potential of combining generative AI sentiment analysis with ensemble learning for more accurate financial forecasting, offering a practical tool for investors and analysts. The results underscore the importance of sentiment features in capturing market dynamics that traditional technical indicators may miss.