When encountering ambiguous or culturally complex terms in the source text, first ask the AI to explain them. If the explanation is accurate, have it restate the meaning in the target language and suggest a few terms that best capture the concept. This approach helps ensure a more accurate and culturally appropriate translation.

The best way to use AI in translation is to consider why you are using AI for translation in the first place. If it is a relatively straightforward text of de-contextualized information - business reports, research results, news articles - then there should be no problem. If you are working toward translations that take up cultural complexity, then don't use artificial intelligence. It both insults the original text-and the person who wrote it- and degrades the text that the reader has to work through. If you value the value of cultural texts, then show that by using humans to connect across language cultures. If you don't care enough to commit to that level of translation, then I don't want to read what you are offering.

It goes without saying that the quality of AI-driven translation is subejct-dependant, that means that informative texts are more likely to be rendered satisfactorily than creative texts.

Having said that, some scholars suggest using post-editing tools such as corpora. Of course this applies in technical (or sector-based) translations.

AI is often mistaken when explaining some cultural issues, but usually pretends to be giving the right information. In my opinion, what usually works is proofreading.

Of course, much depends on the language pair.