<p>Social media has become one of the Internet’s most popular and quickly changing data sources, greatly extending the range of data-driven applications. Disaster management is one of the most important areas of these, where social media insights can be used to inform emergency response and resilience planning in real time. Especially in low-income areas, vulnerable populations are disproportionately affected by natural and man-made disasters, which frequently result in significant social and economic losses. This study provides a thorough review of artificial intelligence (AI)-based disaster management systems that use data from social media. We look at several approaches that have shown promise recently, such as deep learning, machine learning, and statistical models. To classify and compare these techniques, a taxonomy of AI approaches is presented. We also draw attention to the growing significance of multimodal data fusion, which combines text, images, and metadata, in improving the precision and resilience of crisis intelligence systems. We also go over these methods’ advantages and disadvantages, point out unsolved issues, and investigate their wider ramifications in relation to social media analytics. The objective of this review is to promote the creation of more resilient and adaptable AI systems to improve disaster preparedness and response activities by pointing out important research gaps.</p>