AI-enhanced ESG analytics appear to boost equity portfolio performance and resilience compared with traditional ESG ratings, delivering higher returns and lower downside risk in the reported backtests; however, the evidence is associative and hinges on proprietary model choices and backtest assumptions.

Green Intelligence in Finance: Artificial Intelligence-Driven ESG Analytics and Sustainable Investment Performance

Hind Gatoi, Islam Belhaoua, Kashf Akhtar, Miqdad Qadir, Nida Mohammad, Muhammad Ali · Fetched March 17, 2026 · Inverge Journal of Social Sciences

semantic_scholar correlational low evidence 7/10 relevance DOI Source PDF

Portfolios built using AI-enhanced ESG signals outperformed both AI-based low-ESG portfolios and portfolios based on conventional ESG ratings on mean returns, Sharpe ratios, and downside-risk measures in the reported backtests, and AI-derived scores show a stronger association with excess returns in regressions.

This study examined the role of artificial intelligence (AI)-driven Environmental, Social and Governance (ESG) analytics in enhancing sustainable investment performance. While traditional ESG ratings had been widely used in responsible investment strategies, they often suffered from data inconsistency, subjectivity and limited coverage of unstructured sustainability information. AI-based ESG systems were increasingly applied to extract deeper sustainability signals from corporate disclosures, reports and external data sources. Using portfolio-level analysis, this study compared the financial outcomes of portfolios constructed using AI-driven ESG indicators with those based on conventional ESG ratings. The results showed that AI-enhanced high-ESG portfolios achieved higher mean returns and superior Sharpe ratios than both AI-based low-ESG portfolios and traditionally rated ESG portfolios. In addition, AI-driven high-ESG portfolios demonstrated lower downside-risk exposure and smaller maximum drawdowns during market stress, indicating stronger resilience. Regression analysis further revealed that AI-derived ESG scores were more strongly associated with excess returns than traditional ESG metrics. These findings suggested that AI improved the informational efficiency of ESG assessment by capturing more accurate, forward-looking sustainability risks and opportunities. The study concluded that AI-driven ESG analytics strengthened the financial relevance of sustainability integration and supported better-informed investment decision-making. The results carried important implications for investors, regulators and corporations seeking to align AI deployment with high-integrity sustainable finance practices, while also highlighting the need for ethical and transparent AI governance in financial markets. References Albuquerque, R., Koskinen, Y., Yang, S., & Zhang, C. (2020). Resiliency of environmental and social stocks: An analysis of the exogenous COVID-19 market crash. The Review of Corporate Finance Studies, 9(3), 593–621. https://doi.org/10.1093/rcfs/cfaa011 Ali, A., & Rafiq-uz-Zaman, M. (2025). Institutional inertia vs. ethical innovation: A comparative analysis of AI governance at The Islamia University of Bahawalpur and Cambridge University Press. Contemporary Journal of Social Science Review, 3(4), 91–102. https://doi.org/10.63878/cjssr.v3i4.1695 Alshareef, M. N. (2025). Artificial intelligence-enhanced environmental, social, and governance disclosure quality and financial performance nexus in Saudi listed companies under Vision 2030. Sustainability, 17(16), Article 7421. https://doi.org/10.3390/su17167421 Bai, J., & Ng, S. (2023). Machine learning and finance. Journal of Economic Perspectives, 37(2), 195–216. https://doi.org/10.1257/jep.37.2.195 Berg, F., Kölbel, J., & Rigobon, R. (2022). Aggregate confusion: The divergence of ESG ratings. Review of Finance, 26(6), 1315–1344. https://doi.org/10.1093/rof/rfac033 Broadstock, D. C., Chan, K., Cheng, L. T. W., & Wang, X. (2021). The role of ESG performance during times of financial crisis: Evidence from COVID-19 in China. Finance Research Letters, 38, Article 101716. https://doi.org/10.1016/j.frl.2020.101716 Cao, J., Li, J., Li, W., & Zhou, Y. (2023). ESG, big data and investment decision-making. Journal of International Financial Markets, Institutions and Money, 85, Article 101757. https://doi.org/10.1016/j.intfin.2023.101757 Carlei, V. (2025). Outperforming ESG stocks portfolios: A machine learning approach to generate alpha on the S&P 500. Expert Systems with Applications, 260, Article 120157. https://doi.org/10.1016/j.eswa.2025.120157 Chen, M. A., Wu, Q., & Yang, B. (2019). How valuable is FinTech innovation? Review of Financial Studies, 32(5), 2062–2106. https://doi.org/10.1093/rfs/hhy130 Christensen, D. M., Hail, L., & Leuz, C. (2021). Mandatory CSR and sustainability reporting: Economic analysis and literature review. Review of Accounting Studies, 26(3), 1176–1248. https://doi.org/10.1007/s11142-021-09609-5 Dorfleitner, G., Halbritter, G., & Nguyen, M. (2022). The risk-return-ESG profile of sustainable mutual funds. Journal of Asset Management, 23(3), 177–192. https://doi.org/10.1057/s41260-021-00229-3 Drempetic, S., Klein, C., & Zwergel, B. (2020). The influence of firm size on the ESG score: Corporate sustainability ratings under review. Journal of Business Ethics, 167(2), 333–360. https://doi.org/10.1007/s10551-019-04164-1 Fatemi, A., Glaum, M., & Kaiser, S. (2018). ESG performance and firm value: The moderating role of disclosure. Global Finance Journal, 38, 45–64. https://doi.org/10.1016/j.gfj.2017.03.001 Fernando, G. D., Tripathy, A., & Truong, C. (2022). Corporate sustainability and firm performance: Evidence from global firms. Journal of Business Research, 153, 38–52. https://doi.org/10.1016/j.jbusres.2022.08.011 Friede, G., Busch, T., & Bassen, A. (2015). ESG and financial performance: Aggregated evidence from more than 2,000 empirical studies. Journal of Sustainable Finance & Investment, 5(4), 210–233. https://doi.org/10.1080/20430795.2015.1118917 Gibson Brandon, R., Krueger, P., & Schmidt, P. S. (2021). ESG rating disagreement and stock returns. Financial Analysts Journal, 77(4), 104–127. https://doi.org/10.1080/0015198X.2021.1950807 Grewal, J., Hauptmann, C., & Serafeim, G. (2023). Material sustainability information and stock price informativeness. Journal of Accounting and Economics, 75(1), Article 101430. https://doi.org/10.1287/mnsc.2021.4302 Hamdouni, A. (2025). The role of artificial intelligence in enhancing ESG performance. Journal of Risk and Financial Management, 18(10), Article 572. https://doi.org/10.3390/jrfm18100572 Huang, D., Huang, J., & Wan, J. (2023). ESG investment: A review of the literature. Pacific-Basin Finance Journal, 76, Article 102047. https://doi.org/10.1016/j.pacfin.2022.102047 Huang, R., & Renneboog, L. (2020). The effects of ESG ratings on firm performance. European Corporate Governance Institute - Finance Working Paper No. 722/2020. https://doi.org/10.2139/ssrn.3722732 Krueger, P., Sautner, Z., & Starks, L. T. (2020). The importance of climate risks for institutional investors. Review of Financial Studies, 33(3), 1067–1111. https://doi.org/10.1093/rfs/hhz137 Kumar, S., Managi, S., & Matsuda, A. (2022). Stock market response to COVID-19: Evidence from sustainability indices. Economic Analysis and Policy, 74, 44–58. https://doi.org/10.1016/j.eap.2022.01.004 Li, F., Mai, F., Shen, R., & Yan, X. (2020). Measuring corporate culture using machine learning. Review of Financial Studies, 33(5), 2326–2376. https://doi.org/10.1093/rfs/hhz080 Li, Z., Liao, G., & Wang, Z. (2023). ESG performance and financial performance: Evidence from global firms. Technological Forecasting and Social Change, 189, Article 122337. https://doi.org/10.1016/j.techfore.2023.122337 Lim, T. (2024). Environmental, social, and governance (ESG) and artificial intelligence in finance: State-of-the-art and research takeaways. Artificial Intelligence Review, 57(4), Article 76. https://doi.org/10.1007/s10462-024-10708-3 Liu, Y., Song, J., Zhou, B., & Liu, J. (2025). Artificial intelligence applications and corporate ESG performance. International Review of Economics & Finance, 104, Article 104559. https://doi.org/10.1016/j.iref.2025.104559 Pedersen, L. H., Fitzgibbons, S., & Pomorski, L. (2021). Responsible investing: The ESG-efficient frontier. Journal of Financial Economics, 142(2), 572–597. https://doi.org/10.1016/j.jfineco.2020.11.001 Rafiq-uz-Zaman, M. (2022). Redesign for 21st-century skills: Preparing learners for a rapidly changing workforce. Journal of Business Insight and Innovation, 1(2), 89–102. https://doi.org/10.5281/zenodo.17844804 Rafiq-uz-Zaman, M. (2023). Teacher training needs for skill-based education: A review of competencies, barriers, and professional development gaps. Inverge Journal of Social Sciences, 2(3), 166–182. https://doi.org/10.63544/ijss.v2i3.212 Raghunandan, A., & Rajgopal, S. (2022). Do ESG funds make stakeholder-friendly investments? Review of Accounting Studies, 27(3), 822–863. https://doi.org/10.1007/s11142-021-09636-y Walz, U., & Wranik, P. (2023). Artificial intelligence in finance: Applications, risks and regulation. Journal of Banking Regulation, 24(1), 125–143. https://doi.org/10.1057/s41261-022-00194-5 Whelan, T., Atz, U., Van Holt, T., & Clark, C. (2021). ESG and financial performance: Uncovering the relationship. Journal of Sustainable Finance & Investment, 11(4), 311–330. https://doi.org/10.1080/20430795.2021.1920510 Widyawati, L. (2020). A systematic literature review of socially responsible investment and environmental social governance metrics. Business Strategy and the Environment, 29(2), 619–637. https://doi.org/10.1002/bse.2393 Wong, W., & Zhang, X. (2022). Is ESG investing a fad or the future? Evidence from global stock markets. Journal of International Financial Markets, Institutions and Money, 80, Article 101618. https://doi.org/10.1016/j.intfin.2022.101618 Xiao, Y., & Xiao, L. (2025). The impact of artificial intelligence-driven ESG practices on sustainable development performance of Chinese central state-owned enterprises. Scientific Reports, 15, Article 8548. https://doi.org/10.1038/s41598-025-17046-6

Summary

Main Finding

AI-driven ESG analytics produce more informative and stable sustainability scores than conventional ESG ratings, and portfolios constructed using AI-enhanced high-ESG signals exhibit better risk‑adjusted returns and resilience. Specifically, AI-enhanced high-ESG portfolios achieved higher mean returns and superior Sharpe ratios, lower downside risk and smaller maximum drawdowns during stress periods; regression tests show AI-derived ESG scores are more strongly associated with excess returns than traditional ESG metrics.

Source: Gatoi et al., "Green Intelligence in Finance: Artificial Intelligence‑Driven ESG Analytics and Sustainable Investment Performance", Inverge Journal of Social Sciences (Vol. 5, Issue 1, 2026). DOI: https://doi.org/10.63544/ijss.v5i1.216

Key Points

AI vs traditional ESG
- AI-derived ESG scores had a higher mean (72.84) than traditional ESG ratings (65.47) and lower dispersion (SD 9.62 vs 12.35), suggesting greater consistency and coverage of unstructured disclosures.
- Authors argue AI can extract forward‑looking sustainability signals from textual disclosures, media and alternative data that conventional ratings miss.
Portfolio performance
- Portfolios formed on AI-enhanced high-ESG signals outperformed both AI-based low-ESG portfolios and conventional ESG portfolios on average returns and Sharpe ratio.
- AI-based high-ESG portfolios showed lower downside-risk metrics and smaller maximum drawdowns in stressed market episodes, indicating higher resilience.
Statistical evidence
- Regression analyses (controls: firm size, leverage, industry, region) indicate AI-derived ESG scores have a stronger and more robust association with excess returns than traditional ESG scores.
Risks and caveats flagged by the authors
- Potential algorithmic bias, opacity of AI models, and reliance on noisy or impression-managed disclosures can contaminate AI outputs.
- The environmental footprint of AI systems and the need to align AI infrastructure with ESG principles (i.e., "sustainable AI") are highlighted.
- Concerns about transparency, governance and greenwashing remain even with AI tools.

Data & Methods

Design: Quantitative, longitudinal portfolio-level analysis using secondary data.
Population/sample: Publicly listed firms in major equity indices worldwide; purposive sampling to include firms with complete ESG and financial data (authors do not report an explicit sample size in the provided excerpt).
Data sources: Specialist ESG providers (including AI‑augmented ESG analytics) and standard financial databases; harmonised identifiers and cleaned time series.
Key variables:
- Independent: AI-driven ESG analytics (AI-derived ESG scores) vs traditional ESG ratings.
- Dependent: Sustainable investment performance measured by mean returns, Sharpe ratio (risk‑adjusted return), downside-risk measures and maximum drawdown.
- Controls: Firm size, leverage, industry classification, geographic region.
Portfolio construction:
- Firms sorted into high‑ESG and low‑ESG portfolios separately for AI-derived and traditional metrics; historical performance compared across portfolios.
Analyses:
- Descriptive statistics (score means, SDs), portfolio performance comparisons (returns, Sharpe, volatility, downside measures), regression models linking ESG measures to excess returns, and robustness checks including performance during market stress.
Limitations noted or inferable from paper:
- Purposive sampling may introduce selection bias; published excerpt lacks explicit sample size, calendar period and exact return/drawdown figures.
- Possible survivorship bias and unobserved confounders (endogeneity between ESG signals, investment flows and returns) are not fully described in the excerpt.
- Full transparency about the specific AI models, training data, feature engineering and explainability measures is not reported in the provided text.

Implications for AI Economics

Informational efficiency and alpha
- AI-based ESG analytics can increase the informational content of sustainability signals and appear to generate economically meaningful improvements in portfolio performance. This suggests potential for persistent alpha where AI uncovers underpriced sustainability risk/opportunity not captured by legacy ratings.
Market structure and adoption effects
- Widespread use of AI ESG signals could reallocate capital toward genuinely sustainable firms, changing price dynamics and potentially compressing the return premium as signals are arbitraged away. Large asset managers with resources to build/validate AI pipelines may gain competitive advantage.
Externalities and measurement of true sustainability
- Evaluating AI-generated ESG benefits requires accounting for the environmental and social footprint of the AI systems themselves (compute energy, data sourcing). Net welfare claims should be energy‑adjusted.
Regulatory and governance considerations
- Findings strengthen the case for regulatory standards on ESG data, AI model transparency, auditability, and disclosure of model inputs/limitations to reduce greenwashing and algorithmic bias.
Research and policy priorities
- Need for causal identification (instrumental variables, natural experiments) to separate signal discovery from endogenous investment flows.
- Comparative studies across multiple AI providers and across geographies/time to assess robustness and generalisability.
- Development of metrics that combine predictive power with explainability and energy‑efficiency to guide policy and investment practice.

Suggested next research steps (for AI economists and policy makers) - Test for causality: use event studies or exogenous shocks to see whether AI ESG signals predict returns or simply reflect contemporaneous information and flows. - Compare multiple AI providers and model types to quantify model risk and provider concentration effects. - Measure energy‑adjusted risk‑adjusted returns to assess the net climate impact of AI-enabled sustainable investing. - Explore market‑level feedbacks: how AI-generated scores change firm behaviour, disclosure incentives and market prices over time.

If you want, I can: (a) extract and tabulate the study's reported portfolio performance metrics (means, Sharpe, drawdowns) if you can provide the missing tables or pages; or (b) draft suggestions for an econometric strategy to identify causal effects of AI ESG scoring on returns.

Assessment

Paper Typecorrelational Evidence Strengthlow — The analysis is observational/portfolio-backtest based and reports associations between AI-derived ESG scores and subsequent returns; there is no clear exogenous variation or quasi-experimental design to isolate causal effects. Results are vulnerable to selection/survivorship bias, data‑snooping/overfitting, look‑ahead bias, omitted variable confounding (e.g., exposure to common risk factors), and limited reporting of robustness (transaction costs, out-of-sample validation, multiple-testing adjustments). Methods Rigormedium — The study uses standard finance tools (portfolio sorts, Sharpe ratios, downside-risk metrics, and regressions of excess returns) which are appropriate for the question, but the write-up omits crucial methodological details needed to assess rigor: precise sample period and universe, how AI models were trained and validated, controls for known risk factors, treatment of transaction costs and liquidity, and robustness checks (out-of-sample tests, holdout periods, alternative specifications). Those omissions limit confidence even though the employed techniques are standard. SamplePortfolio-level analysis of equity portfolios constructed using AI-derived ESG indicators versus portfolios based on conventional ESG ratings; AI signals were extracted from corporate disclosures, reports and external unstructured data sources; comparisons report mean returns, Sharpe ratios, downside risk and maximum drawdowns, and regressions relate AI-derived ESG scores to excess returns (exact sample period, geographic coverage, and asset-universe not specified in the summary). Themesadoption innovation GeneralizabilityLikely limited to the specific equity universe and sample period used (period not specified); results may not hold for other asset classes (bonds, private equity) or markets., Proprietary AI models and training data are not fully described, reducing replicability and transferability to other AI systems., Backtest and portfolio construction choices (survivorship bias, rebalance frequency, transaction costs, capacity limits) can materially affect performance and are not fully reported., ESG measurement heterogeneity: results may depend on the chosen conventional ESG provider and the way AI maps unstructured signals to scores., Market regime dependence: resilience during one stress episode may not generalize to different crises or future conditions.

Claims (9)

Claim	Direction	Confidence	Outcome	Details
AI-enhanced high-ESG portfolios achieved higher mean returns and superior Sharpe ratios than both AI-based low-ESG portfolios and traditionally rated ESG portfolios. Firm Revenue	positive	medium	Portfolio mean returns; Sharpe ratio	higher mean returns and superior Sharpe ratios for AI-enhanced high-ESG portfolios 0.09
AI-driven high-ESG portfolios demonstrated lower downside-risk exposure and smaller maximum drawdowns during market stress, indicating stronger resilience. Firm Revenue	positive	medium	Downside-risk exposure; maximum drawdown	lower downside-risk exposure and smaller maximum drawdowns for AI-driven high-ESG portfolios 0.09
Regression analysis revealed that AI-derived ESG scores were more strongly associated with excess returns than traditional ESG metrics. Firm Revenue	positive	medium	Excess returns (dependent variable); strength of association with ESG scores	AI-derived ESG scores more strongly associated with excess returns (regression result) 0.09
AI improved the informational efficiency of ESG assessment by capturing more accurate, forward-looking sustainability risks and opportunities. Decision Quality	positive	low	Informational efficiency of ESG assessment (interpreted, not directly measured in summary)	AI improved informational efficiency of ESG assessment (interpretive claim) 0.04
AI-driven ESG analytics strengthened the financial relevance of sustainability integration and supported better-informed investment decision-making. Decision Quality	positive	low	Financial relevance of sustainability integration (qualitative/conclusion)	AI-driven ESG analytics supported better-informed investment decision-making 0.04
Traditional ESG ratings often suffered from data inconsistency, subjectivity and limited coverage of unstructured sustainability information. Other	negative	high	Quality attributes of traditional ESG ratings: data consistency, subjectivity, coverage of unstructured information	traditional ESG ratings suffer data inconsistency, subjectivity, limited coverage 0.15
AI-based ESG systems are increasingly applied to extract deeper sustainability signals from corporate disclosures, reports and external data sources. Adoption Rate	positive	medium	Adoption/application of AI systems for extracting sustainability signals (descriptive)	AI-based ESG systems increasingly applied to extract sustainability signals 0.09
The results carry important implications for investors, regulators and corporations seeking to align AI deployment with high-integrity sustainable finance practices, and highlight the need for ethical and transparent AI governance in financial markets. Governance And Regulation	mixed	speculative	Policy and governance implications (qualitative/recommendation)	implications for ethical and transparent AI governance in financial markets 0.01
The study used portfolio-level analysis to compare the financial outcomes of portfolios constructed using AI-driven ESG indicators with those based on conventional ESG ratings. Other	null_result	high	Study methodology (portfolio-level comparative analysis)	portfolio-level comparative analysis (method statement) 0.15