Writers worry that AI will flatten their voice. It will, if you let it write the piece. It will not, if you use it for the two jobs it is good at: structuring an idea you already have, and producing a rewrite you then argue with. The word-limit check in this lesson is the tool that keeps it honest.
Task 1: an outline from your idea
| Prompt | Result |
|---|---|
| "Write an outline for an article about remote work" | Five generic headings you have read a hundred times |
| "My argument: remote work failed for junior staff because nobody taught them the unwritten rules. Outline a 1,200-word piece: opening scene, three supporting sections each with one concrete example I should find, one counter-argument, close." | A structure built around your argument, with gaps marked for your reporting |
Your turn
Your turn: the outline
My argument in one sentence: [ARGUMENT]. Outline a [LENGTH]-word [FORMAT] that makes this argument. Structure: an opening scene or hook, [SECTIONS] supporting sections each with one concrete example I should go and find, one section that takes the strongest counter-argument seriously, and a close that returns to the opening. Give headings and one line under each. Do not write the prose.
Hint: "Do not write the prose" keeps the piece yours. You want scaffolding, not a first draft you will spend an hour un-writing.
Task 2: a rewrite in a chosen tone, with a limit
Rewrites are where models over-deliver: ask for "tighter" and you get longer. So set a number. This one asks for a plain, dry rewrite of an over-written paragraph in under 60 words, and the check badge tells you whether it obeyed. Naming the tone precisely ("plain, dry, slightly wry") works better than "better" or "more professional", which the model reads as an invitation to add adjectives.
Try it
Rewrite in a plain, dry tone under 60 words
Knowledge check
You asked for a rewrite under 60 words and got 94, with a note saying it kept the length "to preserve nuance". What is the cheapest fix?