The faithful transfer of contextually-embedded meaning remains one of the most persistent challenges in contemporary machine translation (MT) and is particularly evident when dealing with culture-bound terms—expressions or concepts deeply rooted in specific languages or cultures, resisting direct linguistic transfer. Existing computational approaches to explicitating such terms have focused exclusively on in-text solutions, overlooking paratextual apparatus such as footnotes and endnotes systematically employed by professional translators. In this paper, we formalize Genette (1997)'s theory of paratexts from literary and translation studies to introduce the novel task of paratextual explicitation for MT. We construct a dataset of 560 expert-aligned paratexts from four English translations of the classical Chinese literary collection Liaozhai and evaluate LLMs in implicit and explicit reasoning modes on both choice and content of explicitation. Experiments using three intrinsic prompting and one agentic retrieval method establish the inherent difficulty of this task, with human evaluation showing that LLM-generated paratexts improve audience comprehension 91.7% of the time, but with markedly less effectiveness than translator-authored ones. Our findings demonstrate the potential of paratextual explicitations for cultural mediation and advancing MT beyond surface-level equivalence, with promising extensions to monolingual explanation and personalized adaptation.