The circular economy, explained by AI

Abstract

Research on Artificial Intelligence (AI) and the circular economy (CE) has largely examined how AI can improve circular operations. Much less attention has been paid to AI tools widely used to understand and reason about the CE, particularly large language models (LLMs) such as ChatGPT. Researchers, practitioners, students, and decision-makers use these tools to seek explanations and support reasoning about circularity. This paper examines what is at stake when LLM-generated accounts enter CE inquiry. As a conceptual contribution, it develops a framework for analysing how AI-supported CE knowledge is formed, how it may influence action, and what ethical implications follow. The Content dimension concerns the construction and credibility of CE knowledge, focusing on pattern-based generation, opaque and heterogeneous sources, and fragmentation within CE scholarship. The Transformative dimension considers how such knowledge may influence innovation and entrepreneurship, including whether familiar, optimisation-oriented approaches overshadow systemic alternatives. The Moral dimension addresses the risks of relying on LLM outputs, including operational indifference to truth, AI assertiveness, the absence of moral agency, and the need for human judgement and accountability. The paper argues that LLMs can widen access to CE knowledge and support inquiry, but cannot independently determine which interpretations are better supported, exercise the judgement required to challenge linear assumptions, or decide how environmental, social, and economic values should be balanced. It identifies governance priorities concerning transparency, expert scrutiny, human agency, and epistemic humility. The framework can also guide future empirical research on how LLM-generated accounts vary across models, prompts, languages, and user contexts.