Machine translation (MT) accounts for the largest, and increasingly growing, proportion of translated text produced in the 21st century, yet the most influential theoretical frameworks developed in translation studies had preceded the contemporary MT era by decades. The question arises as to whether these theories are well-equipped to capture the makeup of today's corpus-based MT. The current article examines and compares the explanatory power of different theoretical frameworks with regard to MT processes and phenomena. The article discusses the incongruences between corpus-based MT algorithms and major paradigms of (human) translation such as natural equivalence, Skopos theory, and poststructuralist approaches. The article then argues that Descriptive Translation Studies (DTS), in its classic formulation introduced and advanced by Gideon Toury, is the theoretical approach that best corresponds to, and is the most useful framework for conceptualizing, the features of corpus-based MT – and, consequently, of an overwhelming proportion of the translation produced today for multilingual communication and understanding.