Translation practices could be characterised by some bumpy paths, challenging sections, even insurmountable obstacles, irrespective of language combination one works with. Metaphorical concepts, once found in the texts being translated, can only contribute to the complexity of the whole process. Being cautious with conveying the intended messages, hidden behind abstract language, avoiding calques, integrating culture-bound connotations, and, above all, finding the adequate counterparts in the target language/-es are common guidelines along the way. This paper aims to compare the translation practices of EFL students with those of AI-supported engines when translating English passages, brimming with idiomatic language, into Lithuanian. The decision to choose students over language professionals is based on the aspiration to – obtain an insight into their knowledge of metaphorical language and – to be able, at later stages, to introduce them, as future language teachers and translators to the practices and potential of machine translation. Strengths and weaknesses characterising both sides are discussed, accompanied with some implications pertaining to AI usage in pedagogical contexts.