The COVID-19 pandemic has revealed significant spatial inequalities in disease transmission patterns. This study aims to compare the efficiency of Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR) in the spatio-temporal analysis of factors influencing COVID-19 spread in Kurdistan Province, Iran. In this comparative-analytical study, spatial panel data from 10 counties over 36 months (March 2020–February 2022) were utilized. Monthly data comprised 28 independent variables across five domains (climatic, healthcare, transportation, demographic, and crowded places) and confirmed case counts. Analyses including spatial autocorrelation assessment, adaptive modeling, and model performance comparison using the Akaike Information Criterion (AIC), Adjusted R², and Multiple R² were conducted with ArcGIS Pro and SPSS. Results indicate that GWR performance relative to OLS is not uniform and depends on pandemic phase. The GWR model demonstrated substantial superiority during two peak periods: February 2021 and February 2022. In February 2021, GWR achieved a markedly lower AIC (-114302.57 vs. 135.06 for OLS) and a higher Multiple R² (0.98 vs. 0.91). In February 2022, GWR again outperformed OLS, with an AIC of -103916.79 compared to 140.16 and a Multiple R² of 0.74 versus 0.46. However, during other months (December 2020, February 2021, April 2021, August 2021, and December 2021), the performance difference between the two models was negligible. Climatic factors and transportation infrastructure exerted the greatest influence on disease incidence, with effects varying spatially across the province. While global spatial autocorrelation analysis did not confirm a significant cluster pattern, mapping local GWR coefficients revealed significant spatial non-stationarity in these relationships. The findings indicate that GWR is superior under conditions of strong spatial heterogeneity (i.e., peak months), whereas it performs similarly to OLS under more spatially stable, normal conditions.