The development of a series of tools, combining Artificial Intelligence with structural stability models, advanced remote sensing techniques, and image processing, is proposed for the pre-seismic inspection and the assessment of structural vulnerability of existing buildings in densely built urban centers. The objective is the creation of an active database within a geospatial information system, which can be utilized for intervention planning or prioritization of actions in case of emergencies (civil protection). This paper presents results from an ongoing research project conducted by the authors and showcases image processing applications based on the YOLO object detection software, which is powered by a neural network. First, a YOLOv7 model has been trained to capture geometric façade information and create a geometric digital twin of visible parts of the buildings. AI models, particularly the Language Segment-Anything (Lang-SAM) model, were employed to detect facade elements (number of stories, openings, estimate of structural model). Then, an enhanced YOLOv7 model, trained on a Crack Database [1] was used to detect cracks on building facades through photographs, demonstrating remarkable performance with F1-score and mAP50 values of 0.80 and 0.82, respectively. The obtained information facilitates the automatic filling of relevant aseismic evaluation questionnaires, like FEMA, and can be pipelined with structural analysis and aseismic evaluation software for preliminary and quick investigation of existing structures.