In the realm of solids, metal–organic frameworks (MOFs) offer unique possibilities for the rational engineering of tailored physical properties. These derive from the modular, molecular make-up of MOFs, which allows for the selection and modification of the organic and inorganic building units that construct them. The adaptable properties make MOFs interesting materials for photocatalysis, an area of increasing significance. But the molecular and porous nature of MOFs leaves the field, in some areas, juxtapositioned between semiconductor physics and homogeneous photocatalysis. While descriptors from both fields are applied in tandem, the gap between theory and experiment has widened in some areas, and arguably needs fixing. Here we review where MOFs have been shown to be similar to conventional semiconductors in photocatalysis, and where they have been shown to be more like infinite molecules in solution. We do this from the perspective of band theory, which in the context of photocatalysis, covers both the molecular and nonmolecular principles of relevance.