Using AI to study corporate sustainability messages on social media

What is it about?

Large companies increasingly use social media to communicate about sustainability. These messages are not always explicit: a post may relate to climate action, gender equality, health, clean energy, or decent work without directly mentioning sustainability or the United Nations Sustainable Development Goals (SDGs). We analyze more than 1.3 million corporate tweets and over 700,000 images from Fortune 1000 companies between 2017 and 2022. Because corporate social media is multimodal, combining text and visuals, we use large language models to identify sustainability themes in the text and vision-language models to group images into visual themes, such as environmental scenes, community activities, employees, events, infographics, or industrial infrastructure. The result is a large-scale view of how companies present sustainability online: which topics they emphasize, how this differs across sectors, and how these messages relate to ESG risk and user engagement.

Featured Image

Photo by Artem Beliaikin on Unsplash

Photo by Artem Beliaikin on Unsplash

Why is it important?

Sustainability communication is becoming more visible, but it is also difficult to interpret. Corporate messages are often indirect, selective, and spread across large volumes of text and images. This makes it hard for researchers, investors, consumers, and other stakeholders to see what companies actually emphasize when they talk about sustainability. This study is timely because sustainability communication is becoming more visible, while concerns about selective disclosure and greenwashing are also increasing. Our findings show that companies do not communicate about all sustainability goals equally. Many firms emphasize positive themes such as innovation, economic growth, health, or community support, while more difficult environmental topics may receive less attention in some sectors. The paper has impact beyond the current application because it shows how multimodal AI can be used to analyze large-scale corporate communication without costly manual annotation. This matters because it gives researchers and practitioners a practical way to monitor sustainability messaging across companies, industries, and time. It can help reveal patterns that would be difficult to detect manually, including how higher-risk companies may use particular textual or visual themes to shape public perception.