Principal component analysis (PCA) is increasingly used to derive objective weights for GIS-based groundwater potential mapping, yet the stability of these weights is rarely quantified. This study evaluates the robustness of PCA-derived weights for 16 groundwater-conditioning factors in the Lower Namoi Alluvium, New South Wales, Australia. A baseline PCA was fitted to 100,000 common-valid raster cells, component retention was assessed using the Kaiser criterion and parallel analysis, sampling sensitivity was evaluated across 5,000-200,000 cells, and uncertainty was quantified using 1,000 spatial block-bootstrap resamples. The baseline solution retained five components and explained 74.52% of total variance, whereas parallel analysis supported four components explaining 68.24%. Despite this difference, four- and five-component weights were strongly concordant (Spearman ρ = 0.988); the mean absolute weight difference was 0.284 percentage points, although distance to rivers changed by 1.216 percentage points. Rank agreement with a 500,000-cell reference was already high at 5,000 cells (median ρ = 0.994) and reached 1.000 at 100,000 and 200,000 cells. Spatial block bootstrapping identified a stable leading group comprising clay content, rainfall, temperature, bulk density, and elevation, each appearing among the five highest-weighted factors in 93.1-98.5% of resamples. Internal ordering within this group remained uncertain. The results show that the dominant PCA weighting structure is robust to sample size, but borderline component-retention decisions can affect lower-ranked, component-specific variables. PCA weights should therefore be reported with retention diagnostics and uncertainty intervals rather than as deterministic rankings.