Abstract The integration of artificial intelligence (AI) into sustainable finance has emerged as a pivotal innovation for improving environmental, social, and governance (ESG) evaluation. However, ESG data remain fragmented, inconsistent, and prone to methodological opacity—particularly in climate-related investments. This systematic review analyzes 43 peer-reviewed articles and 23 institutional reports to examine whether machine-learning-based ESG scoring models outperform rule-based systems in predicting sustainable investment performance. Following the PRISMA protocol, we applied thematic coding, regional benchmarking, and methodological evaluation across AI–ESG applications. A comparative framework was used to assess AI’s role in ESG integration, predictive analytics, impact measurement, and risk management, with a focus on the climate domain. We also propose and refine the ESG–AI Maturity Index, evaluating AI adoption capacity across data quality, model transparency, and portfolio integration. Findings reveal that AI-enhanced models—particularly those using ensemble learning and sentiment analysis—demonstrate superior performance in forecasting ESG outcomes. However, challenges such as algorithmic bias, regional data gaps, and lack of interpretability remain persistent. The ESG–AI Maturity Index provides a preliminary diagnostic tool for evaluating institutional readiness to adopt AI in ESG scoring. This review contributes by clarifying the performance advantages and limitations of AI-based ESG systems, proposing a framework for maturity assessment, and highlighting urgent policy and technical needs to advance ethical, scalable AI adoption in sustainable finance.
Elhady et al. (2025) studied this question.