Redshift-space distortions (RSDs) offer an attractive method to measure the growth of cosmic structure on large scales, and combining with the measurement of the cosmic expansion history, they can be used as cosmological tests of gravity. With the advent of future galaxy redshift surveys aiming at precisely measuring the RSD, an accurate modeling of RSD going beyond linear theory is a critical issue in order to detect or disprove small deviations from general relativity (GR). While several improved models of RSD have been recently proposed based on the perturbation theory (PT), the framework of these models heavily relies on GR. Here, we put forward a new PT prescription for RSD in general modified gravity models. As a specific application, we present theoretical predictions of the redshift-space power spectra in the f(R) gravity model, and compare them with N-body simulations. Using the PT template that takes into account the effects of both modifications of gravity and RSD properly, we successfully recover the fiducial model parameter in N-body simulations in an unbiased way. On the other hand, we found it difficult to detect the scale dependence of the growth rate in a model-independent way based on GR templates.