We present results of recurrence analysis of 46 active galactic nuclei (AGN) using light curves from the 70 month catalog of the Swift Burst Alert Telescope (BAT) in the 14-150 keV band. Recent applications of recurrence analysis, an alternative method to Fourier analysis, have shown promise in characterizing the abundantly varied and higher modes of variability and timescales in AGN. Recurrence analysis provides complementary information about the nature of the underlying physics in these objects. We use the surrogate data method to compare all derived recurrence-based quantities of our sample to three sets of stochastic light curves with identical power spectrum, flux distribution, or both, in order to determine the presence of non-linearity and non-stationarity. We compare these quantities with known physical characteristics of each system, such as black hole mass, Eddington ratio, and bolometric and X-ray luminosities, and multi-wavelength astronomical properties (e.g., radio loud vs. quiet, obscured vs. unobscured, beamed vs. un-beamed) and spectroscopic type. We find that almost all AGN in this sample exhibit substantial higher order modes of variability than is contained in the power spectrum, with approximately half exhibiting nonlinear or non-stationary behavior. In particular, we find that Type 1 AGN are more likely to exhibit possible nonlinear or nonstationary behavior versus Type 2 AGN, and radio loud sources are more likely to contain linear deterministic behavior than radio quiet sources.