Human–AI framework reveals design levers for collective landscape stewardship in agriculture

We present a rigorously constructed framework to inform the development of evaluative crite-ria for existing knowledge on collective agri-environmental schemes supporting landscape stewardship in agriculture. It is applied in a systematic review of 96 cases across OECD coun-tries, sourced from 73 peer-reviewed articles. The review connects the analytical power of arti-ficial intelligence in concert with expert domain knowledge. The synthesis of both enables a precise and comprehensive examination of policy and implementation dimensions, yielding critical insights into the schemes’ designs and functioning, and their broader applicability. We find considerable heterogeneity among schemes indicating flexibility for decision-makers to adapt schemes to specific environmental targets at local scale, but that there is still significant room to enhance the role of farmers as key players in collective agri-environmental schemes. By integrating human expertise with the capabilities of large language models, our approach exemplifies how to balance and mitigate the limitations of each, enhancing both the reliability and transparency of the review process.