Copyright (c) 2026 Revista Saberes APUDEP

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Structural failure detection is crucial to ensure the safety and durability of buildings. This study evaluates the capabilities of the ChatGPT-O3 reasoning model in identifying and analyzing structural failures using images. The model was applied to a set of photographs showing common pathologies such as delamination, exposed reinforcing steel, cracks, and moisture. Its results were compared with evaluations performed by human experts, analyzing accuracy, technical consistency, and relevance of the proposed solutions.
ChatGPT-O3 demonstrated a high degree of accuracy (85%) in identifying failures and stood out for its viable solutions aligned with international standards such as ACI 546 and EN 1504 (90% relevance). Recommendations included non-destructive testing, concrete repair, reinforcement protection, and preventive maintenance, showing a comprehensive approach to structural failure management.
Although the model has limitations in adapting to local conditions and complex cases, its results show its potential to complement traditional inspections, reducing subjectivity and speeding up diagnosis times. This study highlights the importance of combining artificial intelligence with expert validation to maximize efficiency in structural monitoring and foster more sustainable practices in civil engineering