You can do it, yes. But use of LLMs is not a binary variable - a merely polished text might look like a generated one in its entirety. So you if you prioritize "human-looking" context (I do not) you'd be in disadvantage.