Three fixes for writing that sounds like AI
The reel named three tells: the not-this-but-that line, the warm-up before the point, and the mic-drop last line. You can fix all three by hand, one post at a time. These three fixes do it for you. Pick one to start. They aren’t a stack you have to run in order.
You’ll leave with one fix you can use on your next draft.
Before you start
Sabrina Ramonov’s is a prompt, so it works in any chat window. The other two are skills, which means you need a tool that supports them, like Claude Code or Codex (humanizer also works in Claude.ai).
If you only have a minute, start with Sabrina’s prompt. There’s nothing to install. If you write in Claude Code or Codex, add one of the other two and run your next draft through it.
Sabrina Ramonov’s humanizer prompt
Credit: Sabrina Ramonov. Open a new chat in ChatGPT, Claude or whatever you use, paste the whole prompt first, then paste your draft or tell it what to write. The last line asks it to check its own answer against the list.
Prompt · Sabrina Ramonov’s humanizer* SHOULD use clear, simple language. * SHOULD be spartan and informative. * SHOULD use short, impactful sentences. * SHOULD use active voice; avoid passive voice. * SHOULD focus on practical, actionable insights. * SHOULD use bullet point lists in social media posts. * SHOULD use data and examples to support claims when possible. * SHOULD use "you" and "your" to directly address the reader. * AVOID using em dashes (—) anywhere in your response. Use only commas, periods, or other standard punctuation. If you need to connect two ideas, use a period or a semicolon, but never an em dash. * AVOID constructions like "...not just this, but also this". * AVOID metaphors and clichés. * AVOID generalizations. * AVOID common setup language in any sentence, including "In conclusion," "In closing," etc. * AVOID output warnings or notes, just the answer. * AVOID unnecessary adjectives and adverbs. * AVOID hashtags. * AVOID semicolons. * AVOID markdown. * AVOID asterisks. * AVOID these words: "can, may, just, that, very, really, probably, basically, could, maybe, esteemed, shed light, craft, craftsman, unlock, discover, skyrocket, abound, revolutionize, disruptive, utilize, illuminate, unveil, pivotal, realm, however, harness, exciting, remarkable, it, remains to be seen, stark, testament, in summary, skyrocketing, opened up" IMPORTANT: Review your response and make sure it matches all the criteria above.
It asks the AI to avoid “not just this, but also this.” It doesn’t name the warm-up or the mic-drop last line.
humanizer, by Siqi Chen
What it does: rewrites AI-written text so it sounds like a person wrote it, without changing what it says. It’s built on Wikipedia’s “Signs of AI writing” page, the guide Wikipedia editors use to spot AI text, and checks for 26 patterns. It doesn’t invent names, numbers or quotes. If a sentence needs a detail you didn’t give it, it asks.
- github.com/blader/humanizer (opens in a new tab)Install steps are in the README.
When to use it: you already have a draft that AI wrote, and you want it rewritten so it reads like a person. Give it two or three paragraphs of your own writing first and it matches your voice.
It lists all three tells: “Not X but Y” first, “A staged run-up” for the warm-up, and “One-line closers” for the mic-drop. It edits for human readers. Getting past AI detectors isn’t a goal, and the README says detectors still flag most of its output.
no-ai-slop, by Peter Yang
What it does: removes 20+ patterns of AI slop from your writing without flattening your voice. Its list starts with binary contrasts (“It’s not X. It’s Y.”), throat-clearing openers (“Here’s the thing”) and fake-profound endings. It lists what it changed.
- github.com/petergyang/no-ai-slop (opens in a new tab)Install steps are in the README.
When to use it: you wrote the draft yourself, maybe with some AI help, and you want the slop out without losing how you sound. Its check mode only points out the patterns and changes nothing, which is handy before you post or when you’re reading someone else’s draft.
It names all three tells: “Binary contrasts,” “Throat-clearing openers” and “Fake-profound endings.”
Sabrina Ramonov, Siqi Chen and Peter Yang made these and shared them. The descriptions above come from their own pages.
The three tells, and which fix names them
This comes from the prompt itself and each README’s own list. Where a list doesn’t name a tell, it says so.
- “It’s not just a tool, it’s a partner.”
- Sabrina’s prompt asks the AI to avoid “not just this, but also this.” humanizer lists it first as “Not X but Y.” no-ai-slop lists it first as “Binary contrasts.”
- The warm-up: “Here’s the thing.” “The truth is.”
- Sabrina’s prompt doesn’t name it. humanizer lists it as “A staged run-up.” no-ai-slop lists it as “Throat-clearing openers.”
- The mic-drop last line
- Sabrina’s prompt doesn’t name it. humanizer lists it as “One-line closers.” no-ai-slop lists it as “Fake-profound endings.”
Keep going when you’re ready.
Pick another guide for the next problem you’re working on.