Landslides threaten lives, property, and infrastructure worldwide, with climate change and urban expansion further increasing these risks. Early detection systems are essential for mitigating the impact of landslides. Traditional methods, such as visual inspections and precipitation analysis, are limited in their scope, making it challenging to monitor large areas comprehensively or detect subtle changes preceding landslides. This paper introduces TerrainGuard, a smart landslide detection and warning system that integrates IoT-based sensors, such as rain gauges, inclinometers, and seismic sensors, with deep-learning algorithms for real-time risk assessment. The system uses sensor fusion and cloud-based processing to collect and analyze environmental data, identifying patterns that indicate potential landslides. Compared to traditional methods, TerrainGuard provides accuracy and efficiency through automatic data extraction and fusion (Indukala in IoT-Driven Microseismic Sensing System and Monitoring Platform for Landslide Detection, IEEE, 2024, Piciullo et al. in Nat Hazards 114:3377–3407, 2022). Alerts can be sent quickly through platforms such as Google Maps, Big Alert, or mobile phone alerts to ensure timely evacuation or mitigation. XAI techniques combine the challenges posed by the “black box” nature of deep learning models to make predictions more intuitive. The introduction of wireless sensor networks (WSN) and LoRaWAN modules can enable real-time monitoring even in remote areas (Saldhi in Development of an Automated Monitoring and Warning System for Landslide Prone Sites. Indian Institute of Technology Delhi, 2021), Calvello in LandAware: A new international network on Landslide Early Warning Systems, 2020. TerrainGuard’s architecture offers a variety of layout options, making it ideal for areas prone to landslides. This solution shows the ability to combine IoT, machine learning, and XAI to develop powerful systems that ensure public safety and reduce economic losses. Future work will explore further optimization through cross-learning and integration with existing monitoring infrastructure.