The difference between editing and rewriting an AI draft

When an AI hands you a draft, the first decision is not whether it is good. It is what you are going to do to it, and there are two very different acts with similar-looking results.

Editing preserves the draft’s decisions and fixes the seams. The structure stays. The argument stays. The order of ideas stays. You cut a word that repeats, you smooth a sentence that trips, you fix a fact that is off, you move one paragraph because it landed a beat early. The draft is still the draft when you are done. You have been its editor.

Rewriting replaces the decisions. You keep the subject, and maybe the best single sentence, and you build the rest again from your own sense of how it should go. The structure changes, the argument shifts, the voice becomes yours rather than the draft’s. When you are done, the original is a scaffold that has been taken down.

The skill is knowing which one the piece needs, and the cost of guessing wrong is real. Edit a draft that needed rewriting and you get a polished version of a wrong idea. Rewrite a draft that only needed editing and you lose work that was mostly right, plus whatever good phrasing was already there.

Let me show the two acts on a small invented example, my own text, not something a model produced for me. The draft sentence: “The point is that the old building feels different after dark, which matters because people see it differently at night.”

Editing keeps it and trims it: “After dark, the old building feels different, and people see it differently.” Same idea, same order, fewer words. That is an edit.

Rewriting changes the decision underneath: “The building changes after dark. The same walls, the same steps, and nothing is the same, and you are about to understand why people hurry past it at night.” Same subject, but the draft now leads with the felt change instead of the abstract point. That is a rewrite, and it is a different piece.

This split is mine, but it sits next to an older distinction. The UNC Writing Center’s handout on revising separates proofreading and rewording from the larger work of rethinking a paper’s ideas, and it says revision does not always mean rewriting the whole thing. That is guidance for a student revising their own paper, and it is not about AI drafts. What it shares with my framing is the core point: surface changes and structural changes are different jobs, and the difference is worth seeing.

How do you choose? Look at whether the draft’s underlying decision was right. If the idea and the order are sound and only the sentences are rough, edit. If the idea is sound but the draft aimed it wrong, or led with the wrong thing, or spoke in the wrong register, rewrite. The edit is about the words. The rewrite is about the decisions, and the decisions are the part worth getting right before you spend any time on the words.