Abstract
Wildland-Urban Interface (WUI) areas are critical zones where human settlements and wildland fuels intersect, resulting in heightened wildfire risk and increased potential for losses. This study presents a two‑phase review of literature where artificial intelligence (AI) is applied in wildfire and WUI studies, combining bibliometric mapping for wildfires and a thematic content analysis for WUI. In Phase One, over 3500 publications indexed in Scopus were analysed, searching generally for AI applications in wildfire research to investigate trends. This analysis revealed four dominant clusters: deep learning around computer vision, remote sensing-driven analysis, Random Forests and ensembles with environmental variables, and decision-support prediction. For this phase, the results suggested a notable rise in machine learning after 2016, potentially driven by increased computational capacity and the availability of high-resolution remote sensing, aligned to concerns about the growing impacts of wildfires. Phase Two focused on 52 peer‑reviewed studies explicitly addressing AI for WUI contexts. They were categorised into five thematic areas: risk and vulnerability, human behaviour and decision support, vegetation and fuel, detection and monitoring, and fire behaviour and spread. Models such as random forests, convolutional neural networks, and traditional statistical integrate various datasets, including satellite and UAV/LiDAR imagery, GIS information, and topographic-climatic and socio-economic indicators. The results indicate that AI has contributed to accurate models in spatial resolution and predictive risk for WUI areas, but faces challenges such as simplified WUI maps, fragmented and scarce data, limited model generalisation, and interpretability. Overall, this review identifies opportunities for detailed and diverse WUI mapping, explainable AI, and interdisciplinary collaboration providing means for actionable strategies for WUI communities.