Arid agricultural systems face increasing sustainability challenges due to water scarcity, climate variability, and structural resource constraints. Although Artificial Intelligence (AI) is widely promoted as a key enabler of sustainable agriculture, empirical evidence on its long-term effects on agriculture-related Sustainable Development Goals (SDGs), particularly in arid regions, remains limited. This study investigates the role of AI in supporting sustainable agricultural development in Saudi Arabia’s Ha’il region. Using annual data from 1995 to 2025, AI adoption—proxied by SDG9 indicators that reflect AI-enabling digital infrastructure and innovation readiness rather than observed on-farm AI deployment—is examined in relation to a composite Sustainable Agricultural Development Goals index (SADGH), which integrates SDG2 (food security), SDG6 (water management), SDG8 (economic performance), SDG12 (responsible production), SDG13 (climate action), and SDG15 (land sustainability). Econometric analysis based on a Vector Error Correction Model (VECM) reveals a stable long-run relationship between AI adoption and agricultural sustainability, with approximately 32% of short-term disequilibrium corrected annually. In the short run, AI adoption is positively associated with food security, economic performance, and land sustainability, while water- and climate-related indicators adjust more gradually. Dynamic analyses suggest that AI-related shocks may generate cumulative effects over time. In addition, deep learning models using Long Short–Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures are applied within an exploratory framework to capture potential nonlinear dynamics and generate indicative forecasts. The GRU model shows lower prediction errors; however, results should be interpreted with caution, given the limited sample size. Overall, the findings suggest that AI may contribute to sustainable agricultural development in arid regions, while highlighting the need for further research based on larger datasets.
Triki et al. (2026) studied this question.