Automated ESG Prediction through Artificial Intelligence: A Literature-Driven Empirical Synthesis and Framework for Future Research

Recent advances in AI have revolutionized ESG analytics by extracting sustainability insights from unstructured financial and news text. This study reviews current methods for ESG prediction using NLP, ensemble learning, and deep time-series models. It highlights the strong performance of transformer-based and hybrid CNN-Transformer architectures, alongside challenges in data bias, rating inconsistency, and interpretability. A framework is proposed for multimodal data fusion, cross-provider calibration, and explainable model design, addressing both ethical and practical aspects of sustainable AI.