Non-linear optical microscopy techniques (NLO), which exploit processes upon which a sample emits light in the visible when excited with infrared photons [1], have attracted important attention to date. Among these, Second Harmonic Generation Microscopy (SHG) and Two-Photon Excited Fluorescence Microscopy (TPEF) have been demonstrated as powerful tools for the 3D visualization of tissues and advanced materials. Here we discuss two architectures for super-resolved non-linear optical microscopy. First, we present how the contrast mechanisms of SHG and TPEF imaging can be harnessed to provide resolutions beyond the diffraction barrier, when combined with the re-scan concept, previously introduced in the context of re-scan confocal microscopy [2]. Considering the current high need for techniques capable to characterize non-fluorescent samples at sub-diffraction resolution, we place special focus on showcasing the resolution advantage of Re-Scan Second Harmonic Generation Microscopy (rSHG) [3]. Second, we turn our attention to super-resolved non-linear optical microscopy based on image scanning microscopy concepts [4],[5]. In the final part we discuss perspectives on combining super-resolved NLO imaging with trending artificial intelligence methods. In this context, we first discuss avenues that we are exploring for virtual super-resolved NLO based on Generative Adversarial Networks, and then switch focus towards automated diagnostics perspectives building on the combined use of complementary Deep Learning models [6].