Artificial intelligence in life cycle assessment: methods, applications and challenges

Artificial Intelligence (AI) is reshaping how Life Cycle Assessment (LCA) is applied in sustainability research and industry practice. This review aims to examine how AI is being applied across the stages of LCA, focusing on its applications, capabilities and key challenges, and on its role as a complement to established LCA methods. Eligible studies were identified through a systematic search covering 2010–2026, using conventional AI and machine-learning terms together with extended terms for large language models (LLMs), generative AI and AI agents. Evidence from 39 eligible studies, all published between 2021 and 2026 and reflecting the recent emergence of AI-enabled LCA research, was synthesised, with twelve studies selected for further detailed comparative analysis based on (i) diversity of the AI methodological approaches used and (ii) their contributions to LCA data quality, modelling capability, and interpretability. Foundational AI and LCA methodology literature was incorporated through backward citation tracking to provide context. Half of the twelve studies selected for detailed comparison applied or reviewed machine-learning or neuro-fuzzy models, including two reviews that synthesise a wider body of ML applications (e.g., 81 studies in Ghoroghi et al., 2022). These studies mainly reported gains in predictive accuracy, which depended on the quality of the training data and on validation, while uncertainty and incomplete inventory data remained recurring limitations. At the same time, ontology-based and digital-twin frameworks can offer greater data interoperability and support dynamic assessment. LLM- and agent-based approaches are emerging as a means of automating data retrieval and wider LCA workflows but remain at an early stage of validation. Persistent challenges include opaque model logic, high computational demand, inconsistent data governance, and limited ethical oversight. Addressing these issues will require transparent modelling protocols, energy-aware computation, and harmonised data standards supported by cross-sector governance. Unlike previous reviews that focus predominantly on machine learning applications, this review provides a comparative synthesis of machine learning, LLM, ontology-based, digital-twin, and agent-based approaches across the full LCA workflow. Our analysis of reviewed literature indicates that AI has significant potential to enhance LCA efficiency, data interoperability, and decision support, while highlighting the need for greater transparency, rigorous validation, and robust governance frameworks to support wider adoption. The review advances the field by proposing a conceptual framework for evaluating the strengths, limitations, scalability, and interpretability of AI applications in LCA, together with practical recommendations for developing accountable, energy-efficient, and policy-relevant sustainability analytics.