Green Industry

Large language model decodes European corporate ESG progress: transparency gap narrows, but social performance stagnates

A study based on large language models systematically analyzed ESG reports from 600 large European companies over a decade, revealing core findings such as a narrowing transparency gap, improvement in environmental indicators, and nearly stagnant social performance (except for gender equality). This study provides a data-driven validation framework for the implementation of the European Sustainability Reporting Standards (ESRS).

When AI Meets ESG: A Data-Driven Decoding of European Corporate Sustainability Truths

European companies are often seen as global pioneers in ESG (Environmental, Social, and Governance) practices, but how do they actually perform? A recent study published in *Nature Communications* provides a systematic answer based on large language models (LLMs). Researchers developed an open-source machine learning framework to automatically extract 2.9 million ESG indicators from the annual and sustainability reports of 600 large European companies from 2014 to 2023, covering the three dimensions of environment, society, and governance. They assessed transparency and performance against the European Sustainability Reporting Standards (ESRS). This is one of the most comprehensive and granular quantitative analyses of European corporate ESG practices to date.

The Transparency Gap: Top-Rated Companies Disclose 22% More, but the Gap Is Narrowing

One of the study's core findings is the "transparency gap"—companies in the top 10% of ESG ratings disclose an average of 22% more ESG indicators than those in the bottom 10%. However, this gap is not fixed: over the ten-year study period, the transparency gap narrowed significantly, indicating that lower-rated companies are accelerating their catch-up with disclosure standards. This is partly attributed to EU regulatory pressure: the gradual implementation of the Corporate Sustainability Reporting Directive (CSRD) and ESRS has forced more companies to include ESG information in formal disclosures. Yet the study also points out a persistent deviation between the "narrative" and "quantitative indicators" in reports—the richness of qualitative disclosure does not always translate into verifiable quantitative performance.

Environmental Performance: Superficial Improvement and Disclosure Bubbles

The environmental dimension presents a contradictory picture. On one hand, some environmental indicators (e.g., carbon emission intensity per unit of revenue) have decreased, suggesting progress in energy efficiency and cleaner production among European companies. On the other hand, the study captures a critical signal: reported Scope 3 (supply chain) emissions have risen sharply. The researchers point out that this is mainly due to expanded disclosure coverage—more companies are beginning to account for supply chain emissions, rather than actual emission reductions. This means that relying solely on reported data may overestimate corporate climate action effectiveness. The "disclosure bubble" of Scope 3 emissions serves as a warning for investors and policymakers: it is essential to distinguish between improved data availability and genuine performance improvement.

Social Dimension: Gender Equality Advances Alone, While Other Indicators Remain Almost FrozenThe research findings regarding social performance are the most concerning. Among numerous social indicators such as employee diversity, training, health and safety, and human rights, only gender equality shows a continuous improving trend. Other indicators, such as employee turnover rate, occupational accident rate, and supply chain labor rights, have remained almost unchanged over the past decade. This finding challenges the popular narrative that "European companies lead in social awareness." The research team speculates that the lack of clear quantitative targets for social indicators, along with regulatory requirements (such as social reporting standards under CSRD) that are more ambiguous than environmental indicators, are the main reasons for this stagnation.

Methodological Breakthrough: How LLMs Reshape the ESG Data Ecosystem

The core contribution of this study lies in its methodology: using large language models to automatically extract structured ESG data from unstructured text. Traditionally, ESG data has mainly relied on commercial rating agencies (e.g., MSCI, Sustainalytics), which suffer from inconsistent standards, limited coverage, and lagging updates. In contrast, the LLM approach can process vast amounts of reports at low cost and high frequency, directly aligning with regulatory standards (such as ESRS). This provides EU regulators with tools for real-time monitoring of corporate compliance and opens the "interpretable ESG black box" for investors and academics. The study publicly releases the framework and dataset to promote the decentralization and transparency of ESG analysis.

Impact on European Industrial Competitiveness and Policy

From the perspective of the European business environment, the transparency catch-up phenomenon revealed by this study indicates that regulations such as CSRD are reshaping corporate behavior—even the least transparent companies are being forced to increase disclosure. This helps improve the information efficiency of European capital markets and reduce the screening costs of ESG investing. However, the stagnation in social performance is a warning sign: European companies' competitive advantages in environmental technologies (e.g., green energy, circular economy) have not extended to human capital and social inclusion. In the long run, if European companies fail to achieve substantial breakthroughs in social dimensions (e.g., employee upskilling, supply chain human rights management), their "sustainable competitiveness" will face structural imbalances.

Enlightenment for EU policymakers includes: first, the need to strengthen quantifiable targets and audit requirements for social indicators to prevent them from becoming "soft clauses"; second, regulatory attention should focus on standardized calculation methods for Scope 3 emissions to avoid disclosure inflation masking actual emissions; third, the use of AI tools for dynamic verification of ESG reports should be encouraged, reducing reliance on single commercial rating agencies.

Conclusion

Large language models provide an unprecedented "X-ray" of the current state of ESG implementation in European companies. Transparency is improving, environmental performance is making targeted breakthroughs, but the social dimension shows systemic inertia. This reminds us: ESG is not just a reporting game, but should become a core driving force of corporate strategy. In the context of the EU's pursuit of strategic autonomy and the Green Industrial Plan, only by elevating sustainable development from "disclosure compliance" to "performance competition" can Europe maintain its leading position in the global industrial transformation.

Reader cross-check · europebusinessreview

europebusinessreview frames this note through Europe Business Review covers European markets, EU policy, corporate strategy, green industry, innovation...; European Markets / Corporate Europe / EU Policy Watch explains the local editorial angle. Source links should be opened before the summary is reused: dates, names and status changes still need checking.

Source URLs

  1. https://www.nature.com/articles/s41467-026-75160-zPrimary

Related articles

Back to channel