Using AI data systems to extract information from corporate reports, the researchers tracked over 500 environmental, social, andgovernance indicators (ESG) across 600 publicly listed companies across Europe, creating almost 3 million indicator-levelobservations. The openly accessible dataset reveals the underlying trends in corporate sustainability efforts.

“Our study reveals that the total reported emissions increased by over 400% between 2020 and 2023 alone. However, when weexamined the 15 individual categories that make up value-chain related emissions, such as business travel and product use, none ofthem showed a comparable upward trend.

What did increase sharply was the number of Scope 3 categories companies reported on. As companies measure and disclose moreparts of their value chain, their reported total emissions can rise substantially even without a comparable increase in the underlyingactivities. This shows why greater transparency can sometimes look like deteriorating performance,” says Professor Müller.

Scope 3 emissions typically make up the vast majority of a company’s total carbon footprint and are central to how net-zero pledgesand investor ESG assessments are conducted.

The researchers caution that policymakers, investors, and journalists should treat rising Scope 3 figures with care, distinguishing genuine increases in emissions from improvements in how thoroughly companies are reporting them.

This finding is one output of a broader effort to build an open-source AI framework capable of systematically extracting ESG data,most of which are hidden in lengthy PDF reports, and making it available for public scrutiny.

The study was published in the journal Nature Communications.