What this covers
“Rewrite this” is the worst use of ChatGPT for writing — it produces fluent, generic, often worse prose. The actual workflow that works treats ChatGPT as a thinking partner upstream of drafting (audience, outline, angle) and as a sharp editor downstream of drafting (tightening, voice match, structural critique). The middle — actually writing the words — is yours, and that’s where your voice lives. This guide is for writers, content people, founders, marketers, and anyone shipping written work who’s tired of AI prose that all sounds the same.
Who this is for
- Writers and content marketers who want to use AI without losing their voice.
- Founders and operators writing emails, blog posts, and announcements weekly.
- Non-native English writers who want to upgrade prose without it sounding translated.
- Anyone who’s pasted a draft into ChatGPT, accepted its rewrite, and later cringed at it.
When to reach for it
- Blog posts, newsletters, scripts, reports, long emails — anything that goes through drafting and revision.
- When you’re stuck staring at a blank page (outline help).
- When a draft is “done but mid” and you need a sharper editor than your tired self.
- When you need 3 candidate versions of a headline or opening line.
The four stages of useful AI writing help
- Pre-writing: audience, angle, outline. ChatGPT is great here; you’re not committed to any prose yet.
- Drafting: write it yourself. The model produces the same arc you’ve seen 10 times today; that’s the problem.
- Editing: let ChatGPT critique structure, point out weak transitions, suggest cuts. Best stage for AI.
- Final polish: voice match, hook tightening, headline variants. Use cautiously; voice gets sanded off here.
Before you start
- Have a target audience and goal in mind before opening the chat. “Audience: indie devs. Goal: convince them to try our SDK in 5 minutes” gives the model traction.
- Bring at least one writing sample of yours into the chat as voice anchor. ChatGPT defaults to a generic professional voice; your sample is the counterweight.
- Decide format up front (long form, email, thread, etc.). Different formats need different prompts.
- Set a constraint: word count, no jargon list, banned phrases (“we’re excited to announce” is everyone’s least favorite).
Step by step
-
Tell ChatGPT the audience and the goal of the piece in one paragraph. Be specific:
Audience: senior PMs at B2B SaaS companies, 5+ years experience. Goal: convince them to try our analytics tool. They've heard pitches like this 100 times this year. -
Ask for an outline first — 4-6 sections with a one-sentence purpose each. Edit the outline ruthlessly. The outline is where the piece lives or dies.
-
Draft section by section, yourself. Use ChatGPT only when stuck — and when you do, paste your context, your draft so far, and ask for the specific thing (a transition, a stronger opener, a counter-example).
-
After the draft is complete, ask for structural critique only:
Critique this draft for structure: where does it drag, where is the argument weakest, what should I cut? Do NOT rewrite anything. -
Make YOUR cuts based on the critique.
-
Final polish pass: ask for 3 headline variants, 3 opening-line variants, and a “what would I lose by cutting paragraph X” probe.
Prompts that beat “rewrite this”
TIGHTEN
"Read paragraph 2. Suggest 3 specific cuts that would tighten it
without losing meaning. Show me what would remain."
VOICE-MATCH
"Below is one paragraph of mine. Edit this draft to match its voice —
specifically: sentence length, vocabulary register, use of contractions."
STRESS TEST
"What's the strongest objection a skeptical reader would have to
section 3? Where in section 3 is the argument weakest?"Quality check
- Read the draft aloud. Anywhere you slow down or stumble, edit. The model can’t hear you.
- Is the opening line specific to YOUR piece, or could it open 100 other articles?
- Are the claims your reader could actually call you on? Vague = AI; specific = you.
- Did any of your weird, slightly-off phrasings survive? Those are the bits that make a piece sound human.
How to reuse this workflow
- Build a “voice anchor” file with 5-10 paragraphs you’d want to write like. Paste at the start of every editing session.
- Save the structural critique prompt — it’s the highest-leverage one and reusable across any draft.
- Track which AI suggestions you accept vs reject. After 10 pieces, you’ll see a pattern; tune the prompt.
Recommended workflow
Audience + goal → outline → revise outline → draft sections one-by-one (you) → structural critique (ChatGPT) → your edits → headline + opener variants → final read-aloud pass.
Common mistakes
- Skipping the outline step. The piece will meander because the model has nothing to anchor on.
- Asking for the full piece in one shot. You’ll get generic AI prose with no surprising sentences.
- Not specifying audience. “Write a blog post about X” gives you the median article on X — already written 1000 times.
- Accepting the rewrite wholesale. The model averages toward boring. You’re the variance.
- Editing in the chat. Copy your draft to a real editor; the chat box is for short prompts, not 1500-word documents.
- Asking it to “make it more engaging.” You’ll get exclamation marks and emoji.
FAQ
- Does AI writing detection catch ChatGPT-edited prose?: Imperfectly. If you’ve kept your voice and structure, detection rates are low. If you accepted full rewrites, higher.
- Should I use Claude instead?: Claude tends to produce slightly more thoughtful prose; ChatGPT is faster for iteration. Try the same prompt in both and pick the better output.
- Can I write a whole book this way?: A book, no — the model loses the macro arc. Chapter outlines and per-section editing, yes.
- What about voice mode for drafting?: Useful for outlining and brainstorming aloud. Not useful for the actual sentence-level work.
- Is it cheating?: Depends on the venue. School: read the policy. Professional: as long as the ideas are yours and the words pass your bar, the workflow is fine.