The baryonic acoustic signature in the large-scale clustering pattern of galaxies has been detected in the two-point correlation function. Its precise spatial scale has been forwarded as a rigid-rod ruler test for the space-time geometry, and hence as a probe for tracking the evolution of dark energy. Percent-level shifts in the measured position can bias such a test and erode its power to constrain cosmology. This paper addresses some of the systematic effects that might induce shifts; namely, nonlinear corrections from matter evolution, redshift space distortions, and biasing. We tackle these questions through analytic methods and through a large battery of numerical simulations, with total volume of the order ∼ 100 [Gpc3h-3]. A toy-model calculation shows that if the nonlinear corrections simply smooth the acoustic peak, then this gives rise to an ‘‘apparent’’ shifting to smaller scales. However if tilts in the broadband power spectrum are induced then this gives rise to more pernicious ‘‘physical’’ shifts. Our numerical simulations show evidence of both: in real space and at z = 0, for the dark matter we find percent-level shifts; for haloes the shifts depend on halo mass, with larger shifts being found for the most biased samples, up to 3%. From our analysis we find that physical shifts are greater than ∼0.4% at z = 0 for a LCDM model with σ8 = 0.9. In redshift space these effects are exacerbated, but at higher redshifts are alleviated. We develop an analytical model to understand this, based on solutions to the pair conservation equation using characteristic curves. When combined with modeling of pairwise velocities the model reproduces the main trends found in the data. The model may also help to unbias the acoustic peak.