The Event Horizon Telescope (EHT) Collaboration has recently produced the first resolved images of the central supermassive black hole in the giant elliptical galaxy M87. Here we report on tests of the consistency and accuracy of the general relativistic radiative transfer codes used within the collaboration to model M87* and Sgr A*. We compare and evaluate (1) deflection angles for equatorial null geodesics in a Kerr spacetime; (2) images calculated from a series of simple, parameterized matter distributions in the Kerr metric using simplified emissivities and absorptivities; (3) for a subset of codes, images calculated from general relativistic magnetohydrodynamics simulations using different realistic synchrotron emissivities and absorptivities; (4) observables for the 2017 configuration of EHT, including visibility amplitudes and closure phases. The error in total flux is of order 1% when the codes are run with production numerical parameters. The dominant source of discrepancies for small camera distances is the location and detailed setup of the software "camera" that each code uses to produce synthetic images. We find that when numerical parameters are suitably chosen and the camera is sufficiently far away the images converge and that for given transfer coefficients, numerical uncertainties are unlikely to limit parameter estimation for the current generation of EHT observations. The purpose of this paper is to describe a verification and comparison of EHT radiative transfer codes. It is not to verify EHT models more generally.