Neural machine translators (NMTs), such as "Google Translate," may assist second language (L2) readers with general comprehension. However, previous empirical studies show mixed results regarding their effectiveness. In this study, 145 Korean English learners from a girls' high school were asked to solve three types of reading comprehension problems ("grammar judgment, inferring meaning from context, inferring main idea") under three reading conditions (no aid, MT, glossary). Overall, when using MT, reading comprehension scores were higher than in either the no aid or glossary conditions individually. However, none of the reading aid conditions improved "grammar judgment." Only mid-proficiency learners benefited from MT in both "inferring meaning from context" and "inferring main idea" tasks. The results suggest that the glossary may have interrupted the flow of the reading process. With the widespread availability of MT as an online reference tool, L2 teachers should consider incorporating MT as a legitimate reading aid for different proficiency levels and reading purposes.