Last Verified: October 25, 2026 — Quick-scan cheat sheet: copy a prompt, fill the brackets, go; sample runs are real Claude output, not invented
AI writing usually gets called out for rhythm and structure, not word choice: the same three-part lists, the hedge-then-hedge-again sentences, the tidy “in conclusion” wrap-up. Swapping a few words doesn’t fix that. These prompts target the structure.
TL;DR
- Run these as edits on a draft you already have, one or two at a time.
- Prompt 5 (voice sample) has the biggest effect. Feed it something you actually wrote.
- Finish with the read-aloud test (Prompt 6).
On this page (8 sections)
Cheat Sheet at a Glance
| Tell | What it looks like | Prompt |
|---|---|---|
| List-itis | Everything in bullets or threes | #1 |
| Fence-sitting | “On one hand… on the other” | #2 |
| Flat rhythm | Every sentence the same length | #3 |
| Throat-clearing | “In today’s fast-paced world…” | #4 |
| No voice | Could be anyone’s writing | #5 |
The Prompts
1. Ban the list-itis
Rewrite this as flowing paragraphs, not bullet points or numbered lists, unless a list is genuinely the clearest way to present it. Avoid grouping things in threes: [paste text]Sample run · Claude, October 2026. Your wording will differ; the structure should not.
Input we used: Our new onboarding flow offers several benefits: faster setup time, fewer support tickets, and higher activation rates. These improvements result from three key changes we made to the signup process.
Our new onboarding flow cuts setup time, and that one change is dragging two other numbers with it: support tickets are down, because people aren’t getting stuck in the same three spots anymore, and activation is up, because more people finish the flow instead of abandoning it halfway. All three numbers moved for the same reason — we fixed the three places people used to get stuck in signup.
The three-item list didn’t disappear, it got folded into one sentence that shows how the items connect (one fix causing all three). That’s the actual test: can you still find the three things in there, just no longer presented as a list?
2. Force one clear position
Take a clear position on this instead of presenting both sides equally. State your recommendation in the first sentence, then give the strongest reason: [paste topic or draft]Sample run · Claude, October 2026. Your wording will differ; the structure should not.
Input we used: Should our team switch from Slack to Microsoft Teams? Both have strengths. Slack has better integrations, while Teams is included in our Microsoft 365 plan. It really depends on your needs.
Stay on Slack. Your team’s daily work runs through integrations, and that is where Slack is clearly stronger; switching means rebuilding the alerts, bots, and workflows you already rely on. Teams being included in Microsoft 365 is a real saving, but it only matters if the Slack bill is a problem today. Revisit this if your next Slack renewal jumps in price.
It picked a side in the first two words, and it named the condition that would change the answer. Notice the reasoning leans on an assumption (that your team uses a lot of integrations). That’s the sentence to check before you publish.
3. Vary sentence length on purpose
Rewrite this so sentence length varies noticeably. Mix short, punchy sentences with longer ones. No more than two sentences in a row of similar length: [paste text]4. Cut the throat-clearing opener
Delete any opening sentence that doesn't contain new information. Start with the actual point. Also remove any closing sentence that only restates what was already said: [paste text]5. Feed it your real voice
Here's a sample of how I actually write: [paste 200+ words of your own writing]. Match its tone, sentence rhythm, and vocabulary level when you rewrite the following: [paste text]Why it works: A real sample beats any adjective like “conversational”.
6. Run the read-aloud test
Flag any sentence in this that a real person wouldn't naturally say out loud, and rewrite just those. Leave the rest untouched: [paste text]
Worked Example: Two Passes on One Paragraph
Here’s a typical AI-drafted paragraph for a product update:
In today’s fast-paced digital landscape, staying organized is more important than ever. That’s why we’re excited to introduce Saved Views, a powerful new feature designed to streamline your workflow, boost your productivity, and enhance collaboration. Whether you’re a small team or a large enterprise, Saved Views can help you work smarter. In conclusion, Saved Views is a game-changer for teams everywhere.
It has four tells: a throat-clearing opener, a list of three benefits, a both-sides “whether you’re” sentence, and a closing line that only restates. After Prompt 4 (cut the opener and closer) and Prompt 1 (no threes), then Prompt 3 (vary rhythm):
You can now save any filtered view and come back to it in one click. Set up “My open bugs, sorted by age” once, and it’s there every morning. Share a view with your team and everyone sees the same list. No more pasting filter settings into Slack.
Shorter, specific, and it says what the feature actually does. Notice that no clever vocabulary was needed. The fix was all structure: cut what carried no information, and replace abstract benefits with a concrete example.
Make It a Standing Rule
If you fix the same tells in every draft, move the rules upstream so the first draft comes out cleaner. Paste a block like this into ChatGPT’s custom instructions, a Claude Project, or a Claude style:
Writing rules for everything you draft for me:
- Start with the point. No opening sentence without new information.
- No closing sentence that restates what was already said.
- Prefer paragraphs to bullet lists unless I ask for a list.
- Don't group things in threes by default.
- Take a position when I ask for a recommendation.
- Vary sentence length. Short sentences are fine.
- Never use: "in today's fast-paced", "game-changer", "delve", "seamless".A banned-phrase list helps a little, but the structural rules above it do most of the work. A model told only to avoid “delve” will find a synonym and keep the same sentence shape.
For Developers: Docs, READMEs, and Commit Messages
The same tells show up in AI-written documentation: READMEs that open with “This project aims to provide a robust solution”, and commit messages that say “Enhanced functionality and improved code quality”. If you use Claude Code, add a short writing section to your project’s CLAUDE.md:
## Writing
- Commit messages: imperative mood, say what changed and why, no "enhance" or "improve" without specifics.
- Docs: lead with the command or example, then explain. No marketing language.It’s read at the start of every session, so you stop correcting the same phrasing in every review.
Common Mistakes to Avoid
- Asking it to “sound more human”. Too vague. Name the specific tell you want removed.
- Running all six at once. Stacking every instruction produces a new kind of sameness. Apply them in passes.
Q&A
Will this get past AI detectors?
That’s not the goal, and detectors are unreliable. The goal is writing that reads well to people.
Which model is best at this?
Results vary by model and draft. Prompt 5 matters more than the model.
Should I just write it myself?
For anything personal or high-stakes, often yes. These prompts are most useful for drafts you’d otherwise spend an hour polishing: updates, docs, summaries, and first drafts you plan to edit.
Why does my rewrite still sound like AI?
Usually because the source draft had no specifics. No prompt can add a real example, number, or opinion that isn’t there. Add one concrete detail yourself, then run the edits.
Related Guides
Saved you a few rounds of back-and-forth? Follow us on X for more cheat sheets and setup guides.