Cosmologists are struggling to understand why the expansion rate of our Universe is now accelerating. There are two sets of explanations for this remarkable observation: dark energy fills space or General Relativity fails on large scales. We study a rather general class of f(R) theories whose expansion history is compatible with cosmological observations, and show that the growth of structure D is scale-dependent, and systematically enhanced on all scales with respect to a smooth dark energy model with the same expansion history, H(z). We build a null test parameter \epsilon(k,a), and show that it is identically zero on all scales in GR, contrary to the modified gravity case. Such parameter can be written in terms of the combination \Omega_m h^2 probed by the CMB, the linear galaxy bias b, and the linear redshift distortion parameter \beta. We show how redshift surveys in combination with CMB experiments can measure b and \beta, allowing precise determination of \epsilon and thus a precision test of GR. We forecast that an experiment like ADEPT could measure \epsilon to the percent level.