The article "Damage assessment" in Science (Volume 391, Issue 6783, Pages 338-341, January 2026) deals with the assessment of damage caused by disasters.[6] Traditional methods of assessing building damage after natural disasters are time-consuming and expensive, but advances in geoAI (geospatial artificial intelligence) enable automation using computer vision, remote sensing, and machine learning on data from satellites and drones.[1] GeoAI enables faster targeting of recovery resources in affected areas.[1] The study presents deep learning on high-resolution aerial images for automatic detection and classification of structural damage.[1] The SPADANet model achieves 74% of the performance of the fully supervised method with only 10% labeled data and improves the detection of damaged buildings by more than 9% over existing flood models.[2] Challenges include data and model quality, including the need for comprehensive training data capturing different damage types and environmental factors.[1] The review analyzes applications in real disasters by humanitarian organizations and points out lessons learned.[1]