Vegetation structure is a potentially important ecological factor structuring bird communities, but can also affect bird detectability, which can complicate the resolution of ecological patterns. We addressed how vegetation structure may bias the efficiency of the 3 methods most commonly used to sample bird species richness in the Brazilian savanna (Cerrado), which features a gradient of vegetation structure from grasslands to woodlands. We compiled secondary data on Cerrado bird species richness from the scientific literature, master's theses, doctoral dissertations, and regionally important biological publications. We used generalized linear mixed models (GLMM) to analyze the effects of habitat type and sampling method on Cerrado bird species richness estimates, controlling for variation among publications, researchers, and study sites. The data provided strong support for interactive effects between habitat type and sampling method on observed bird species richness. Point count surveys had a greater average number of species sampled per unit time in all habitats, especially in shrublands and grasslands. The number of species sampled per unit time by transect surveys did not vary across the vegetation gradient. Mist net surveys showed a slight decrease in species numbers sampled with increasing vegetation complexity. Random effects accounted for 40% of data variation, mainly because of methodological differences among studies. Despite the amount of work that has been done in the Cerrado, it is not currently possible to ascertain the relationship between bird species richness and vegetation structure because of the interaction between sampling method and vegetation complexity, as well other factors leading to variation among studies. We suggest that estimating detectability is a feasible solution for studies in which the ecological effect of interest also strongly affects sampling, as is the case in the Cerrado.