Zenn guide: rewrite AI drafts into your own verified article

The post starts from a common frustration: an AI draft can have a clean structure yet feel like someone else wrote it. Its diagnosis is that AI supplies the outline while the author holds the facts of what actually happened, so the rewrite is mostly a matter of adding the latter.

The concrete material comes early. Before editing, the author pastes the full terminal log and error messages, such as a build failing with "Cannot find module '@/lib/config'", at the end of the draft. Under each heading, a single personal sentence is inserted ahead of the AI's explanation — the example notes getting stuck for 30 minutes on the paths setting in tsconfig.json. Official-documentation summaries are then cut and replaced with links, since the post argues the value lies in the record of use, not the summary.

Only after that does style come into play. By this point AI and human sentences may be mixed, and the author mentions AI humanizers as one way to smooth stiff phrasing while warning that losing the added specifics would defeat the purpose. Reading the result aloud is the last check. The closing note frames the division of labor simply: AI is good at structure, the author knows what happened.