Modern development of materials with superior characteristics leverages the data-driven design of novel architected structures to target specific mechanical properties. However, this technique requires collecting experimental data, for which generating a certain amount of the desired novel arrangements is necessary. If Laser-Based Powder Bed Fusion (PBF-LB) greatly expanded the field of possibilities for manufactured structures, considerable limitations remain since several local geometrical configurations cannot efficiently be obtained, especially on small scales. This includes, for instance, the presence of internal structures from which the powder cannot be removed. This restriction becomes a real challenge when the structures are highly complex and disordered. This motivates the development of an automated method to spot undesirable geometrical arrangements, select the most printable structures, and efficiently build a sufficient set of representative samples. In this context, algebraic topology offers powerful methods to detect specific local geometrical configurations. We will show how these mathematical concepts can be applied to additive manufacturing in the form of an algorithm that detects the hardly printable features, indicating their positions as well as their size. We will also illustrate the great adaptability of this process by explaining how it can be tuned to detect different kinds of problematic arrangement, without taking any assumption on the considered structure. Finally, we will discuss the implications of this algorithm for the concrete tasks of the data-driven design of metamaterials.