Stimulated (coherent) and spontaneous (incoherent) nonlinear X-ray signals are expressed using a spatially nonlocal response tensor which directly connects them to the time evolving current j and charge σ densities rather than to electric and magnetic multipoles. The relative contributions of the σA(2) and j · A minimal coupling terms, where A is the vector potential, are demonstrated. The two dominate off-resonant and resonant scattering, respectively, and make comparable contributions at near resonant detunings.