A new study suggests that European companies are telling a clearer story about climate impacts—but the details are uneven, especially beyond their own operations.
Using an AI-powered approach, researchers from LMU Munich and the University of Cologne analyzed 2.9 million sustainability-related indicators found in ten years of corporate reporting. The dataset spans the annual and sustainability reports of 600 of Europe’s largest listed companies, covering 2014–2023.
In total, the team reviewed roughly 9,000 PDF documents comprising about 1.7 million pages. The period was chosen to predate the EU Corporate Sustainability Reporting Directive (CSRD), meaning companies were mostly reporting under earlier rules.
To measure what would look like CSRD-aligned transparency, the researchers applied the stricter CSRD disclosure requirements retrospectively. Their method evaluated reporting coverage across environmental, social, and governance (ESG) topics by scanning for 501 defined indicators.
The analysis was conducted with a large language model—Llama-3.1-70B-Instruct—which automatically extracted and classified ESG signals from long-form texts. “Instead of relying on a few costly commercial datasets with inconsistent definitions, we can systematically track what companies actually report,” the authors note.
Results show a steep improvement in disclosure volume: the average number of reported indicators rose by 52.4% from 2014 to 2023. Companies with weaker sustainability performance also narrowed the gap, with reported indicator shortfalls shrinking from 39.4% to 6.8%, suggesting transparency is converging across firms.
But performance signals don’t tell a uniformly reassuring tale. Reported direct emissions declined substantially, while reported indirect value-chain emissions increased more than fivefold—largely because firms are now capturing and publishing more categories rather than necessarily cutting more emissions.
On social metrics, progress is also mixed. Representation of women in top management rose by 9.2 percentage points, while the gap between executive compensation and median employee wages expanded dramatically—more than twelvefold since 2014.
The researchers are releasing their dataset and code through the Sustainability Reporting Navigator open science initiative, aiming to help regulators, investors, and NGOs compare firms more consistently and spot transparency gaps.
Subject of Research: Assessing corporate sustainability disclosure and performance using large language models across ESG indicators.
Article Title: Assessing corporate sustainability with large language models: evidence from Europe
News Publication Date: 7-Jul-2026
Web References: https://doi.org/10.1038/s41467-026-75160-z
References: Nature Communications (DOI: 10.1038/s41467-026-75160-z)
Image Credits:
Keywords: corporate sustainability reporting, CSRD, ESG disclosure, climate data, value chain emissions, large language models, transparency gaps, AI text mining
