PurposeThis study aims to deepen the understanding of what stakeholders talk about when it comes to sustainable fashion on social media and how. Sustainable fashion is a broad umbrella term, which can distract attention from the differences between the individual subtopics and the sentiments ascribed to them. However, little systematic research exists on how the stakeholder activity and dominant sentiments vary across different sustainable fashion topics.Design/methodology/approachThis study is based on a social media analysis of 19,179 tweets authored by 1,819 distinct stakeholders on Twitter (now “X”) from 2007 to 2022. A large language model, a type of artificial intelligence (AI) that focuses on understanding and generating human language, is used to conduct a sentiment analysis of six stakeholder groups and 81 keywords linked to sustainable fashion. Two case examples are used to highlight the differences in stakeholder perceptions of sustainable fashion.FindingsThe social media analysis demonstrates how subcategories of sustainable fashion significantly differ in terms of stakeholder interest, activity and sentiments. For instance, tweets on circular economy and relevant subcategories (closed loop, recycling, upcycling, etc.) are popular, whereas issues linked to environmental, social and governance (ESG) and due diligence receive little attention on social media. While sentiments toward sustainable fashion are in general positive, discussions on topics such as labor rights issues are consistently associated with negative sentiments across most stakeholder groups.Originality/valueThis study contributes to the literature by demonstrating how stakeholders and sentiments vary across different topics linked to sustainable fashion on social media, which has become one of the main channels for communicating sustainability content. The findings thereby shed new light on dominant stakeholder positions regarding a wide variety of sustainable fashion topics.