Zeolites are a class of porous crystalline silicate-based materials with applications such as catalysis and separation. Zeolite intergrowths can have superior performance compared with conventional single-phase zeolites in these applications. This study develops a computational workflow to evaluate ~1.03 trillion atomistic structures to identify promising zeolite intergrowths through geometrical analysis and atomistic simulations. We find that interfacial energy is an excellent descriptor to distinguish hydrothermally synthesized zeolite intergrowths from the others, showing almost-perfect classification performance (area under the curve of 0.995). Computational screening workflow saves 100% of hydrothermally synthesized zeolite pairs and successfully rejects 99.3% of hypothetical pairs. Network analyses reveal that hypothetical pairs comparable to experimentally proven ones show substantial topological and chemical similarities, although such information is not directly used in the screening workflow. One of the hypothetical candidates that passed the criteria is experimentally realized by direct and seed-assisted hydrothermal syntheses, thereby broadening the applicable scope of zeolite intergrowths to zincosilicates with three and nine rings.