Text that sounds human.
UnSlopIt takes anything an AI drafted and strips the tells that give it away. Things we remove without remorse:
- “In today’s fast-paced world…”
- “It’s not just X. It’s Y.”
- “The key takeaway?”
- “Here’s the thing…”
If you used all four today, we’re not judging. Too much.
Paste a draft, pick how aggressively to edit, and get back a clean version plus a red-pen list of every change and why it was made. The method behind it is written up in full: how to humanize AI text. The first three edits are free, with three more once you register your email. Curious what actually counts as slop? Read the field guide →
Editorial
Who’s the editor?
UnSlopIt is written and maintained by one person — me. I write this anonymously; it’s a side project and the arguments should stand without a byline. I’ve spent over a decade in B2B marketing, most of it leading teams that produce a lot of writing: product copy, thought leadership, sales material, executive communications. Reading and editing other people’s drafts for customer clarity, brand voice, and simplicity is a big part of the job. When AI tools arrived, the same handful of patterns started appearing in everything that crossed my desk, and the taxonomy on this site began as a private list of them.
I spend too much time reading LinkedIn, but I also enjoy the classics at bedtime and quality business media, which I mine for their occasional gems.
How was the taxonomy built?
The taxonomy started from patterns observed in AI output. In the beginning, I made the rookie mistake of tracking words, but soon realized it went beyond words. Some of it was words, some of it was patterns, and some of it was the structure of the arguments. I also relied on documented community observations (choice subreddits, published research on overrepresented vocabulary).
I continue to revise the taxonomy as models change and new research gets published. I love doing this stuff. It is almost an Orwellian tradition — not in the 1984 sense but in the ‘Orwell’s 6 Rules of Writing’ sense. In that spirit, take the taxonomy and recommendations from UnSlopIt as broad guidelines and follow Orwell’s Rule #6: break any of these rules sooner than say anything outright barbarous.
When I cite a figure — the share of new pages that are AI, the reader-trust surveys — I keep the full sourcing in one place: the SLOP Equation, where every number links to its primary source.
What data is collected?
This app does not store your drafts or the edits. It records which of the taxonomy categories were detected in your draft, the word count, and when an edit was made — retained indefinitely in aggregate, not linked to any user or account, so I can refine which categories to tighten or expand. For example, ‘engagement bait appeared in 78% of drafts’ would probably signal that the category is too broad.
Corrections policy
What happens when a category is wrong — misattributed, or a statistic turns out to be misreported? Please contact me at editor@unslopit.ai and I’ll do my best to make corrections, and to show all historical corrections so they stay easily visible.
FAQ
- Is my text stored?
- No. Your draft is processed to produce the edit and is never saved — it exists in memory only for the length of the request, and the browser never talks to the AI directly (your text goes to our server, which holds the API key and returns the rewrite). We do log anonymous, non-identifying metadata about each run — which taxonomy categories were flagged, an approximate word count, and a timestamp, never the text itself — and keep those aggregate counts indefinitely so we can cite real statistics about how often each pattern shows up.
- Will it change what I meant?
- No. The editor preserves your meaning, facts, claims, and roughly your length — it never invents new claims or statistics. You choose how hard it edits: Light touch, Medium (default), or Full rewrite.
- How is this different from asking an AI to “rewrite this”?
- A general model often adds its own slop. UnSlopIt runs a purpose-built editor against a fixed taxonomy of tells and shows you exactly what it changed and why — as a redline plus a clean version.
- What's the “add a typo” option?
- Perfectly clean text is itself a tell. Toggle it on and the editor slips in one small, natural typo so the result reads as human-written. We mark where it landed with a spellcheck-style squiggle.
- Can a real person edit it?
- Human editing is on the way. Register your interest on the home page and we'll email you when it's live.
- Is there a limit?
- In the prototype you get a few free edits, then the option to create a free account for more. Drafts are capped at 1,000 words per pass — split longer pieces by section.