As wildfires intensify globally, reliable damage prediction is increasingly vital. This study develops and evaluates Random Forest and XGBoost models to classify structural outcomes (no damage, damaged, destroyed, inaccessible) using pre-fire attributes (e.g., roof type, assessed value, year built). After SMOTE rebalancing, both models achieve ∼90% accuracy, with XGBoost performing slightly better on minority classes. Feature importance analysis reaffirms that newer, high-value homes and fire-resistant construction features (e.g., enclosed eaves, ember-resistant vents) substantially reduce wildfire damage risk. While distinguishing partial damage from destruction remains challenging, the approach markedly enhances risk assessment, guiding targeted resource allocation and evidence-based policies for structural resilience in fire-prone regions.