I imagine you might do well to write your prompt with similar language (diction, sentence construction, etc.) to what you'd find in the output/answer you are at

I imagine you might do well to write your prompt with similar language (diction, sentence construction, etc.) to what you'd find in the output/answer you are attempting to evoke from the LLM. These are prediction machines, after all.