AI-generated text has a fingerprint, and most readers can feel it even when they can’t name it. Flat rhythm, no opinions, em dashes everywhere, sentences that hedge instead of commit. A skill called Unslop targets exactly this problem.
What Unslop Actually Does
Unslop is a prompt-based skill — a block of instructional text you save inside Claude or ChatGPT — that audits a piece of writing for robotic patterns and rewrites them toward something that sounds like a real person had a point of view.
It checks for 31 distinct issues. Some are obvious:
- Overuse of em dashes as a crutch for transitions
- Passive constructions that drain energy from sentences
- Filler openers like “Certainly!” or “Great question!”
Others are subtler. AI tends to avoid the word I even when first-person is the most natural fit. It homogenizes sentence length, producing that relentless medium-length cadence that feels neither punchy nor flowing. It rarely commits to an opinion, preferring “there are benefits and drawbacks” over just saying which one wins.
Unslop flags all of it.
Why This Matters More Than You’d Think
If you’re using AI to draft emails, reports, social posts, or articles, the output is probably serviceable — but it likely reads like it came from a committee. That’s not just an aesthetic problem. Readers disengage. Clients notice. Your voice disappears.
The fix isn’t to stop using AI. It’s to build a cleanup step into your workflow, and Unslop automates that step.
Think about a common scenario: you ask Claude to draft a project update for a client. The draft is accurate and well-organized, but it opens with “This document provides an overview of…” and uses “it is important to note” twice in the second paragraph. Unslop catches both, along with eight other patterns you’d have missed on a quick read.
How to Install It in Claude or ChatGPT
Unslop is free. It’s plain text — no app to install, no API key required.
In Claude
- Find the Unslop skill text (search for it by name; it’s shared openly in the AI community).
- Open Claude and go to your Skills or Projects section.
- Use the skill creator prompt (
/skill creatoror the equivalent in your setup) and paste the Unslop text in. - Save it with the name “Unslop.”
Once saved, you can invoke it any time by typing something like “add Unslop” or “run Unslop on this” before or after your draft.
In ChatGPT
The process is nearly identical. Use the template creator or a custom GPT system prompt to store the skill text. Label it clearly so you can call it by name mid-conversation.
Using It On the Fly
The real power is how low-friction this becomes once it’s saved. You write a draft, then append “Unslop this” at the end of your message. You get back a version with the robotic patterns removed and the rhythm varied. It takes one extra line.
For longer documents, paste section by section rather than the whole thing at once. You’ll get more precise edits and easier review.
What Good Output Looks Like
Before Unslop, a typical AI draft of a product announcement might read:
It is important to note that this update introduces several enhancements that have been designed to improve the overall user experience.
After:
This update makes the app faster to navigate and fixes the three bugs users complained about most.
Shorter. Concrete. No hedge. That’s the difference between text that lands and text that gets skimmed past.
One Honest Caveat
Unslop improves AI-generated drafts significantly, but it’s not a replacement for your own editing eye. It won’t know when a formal tone is actually appropriate, and it can occasionally over-correct, stripping out structure that was there for a good reason. Treat the output as a strong second draft, not a finished product.
The best use of this skill is as a gut-check before you send anything important. Run it, read the result, override anything that feels wrong, and ship.
If you’re writing with AI regularly and not running some version of this check, you’re probably publishing a fingerprint you didn’t mean to leave.