This study investigates the controversy surrounding the sensitivity of masked repetition priming to word frequency: while unmasked priming exhibits a frequency attenuation effect, wherein high frequency words yield smaller repetition effects, this phenomenon has been inconsistently reported in masked priming. We conducted two large online experiments with rigorously validated frequency databases to reconcile past discrepancies. The first experiment confirmed the viability of conducting masked priming experiments in web browser-based settings. The pre-registered second study, designed for high statistical power and precision, identified a 10-ms attenuation effect under masked priming. This result suggests that the repetition effect in masked priming is less qualitatively distinct from unmasked priming than previously assumed. This finding has implications for masked priming experimental design and theoretical consequences for models of priming. Crucially, models that predict either the presence or absence of frequency attenuation under masked conditions need to account for a small but reliable effect.