17% off for the first 250 customers (7 left)

Stop sounding
like ChatGPT—

slopt tracks every linguistic tic, em-dash, and “delve” ChatGPT, Claude, and Gemini are stuffing into your text.

We turn it into custom instructions. Copy & paste into your AI of choice to sound human again.

Stop the slop →Watch the slop live
Loved by 200+ humansPasteable in 30 seconds

What is slop?

The linguistic fingerprints AI leaves on every paragraph. Hover any highlight to inspect the pattern.

In today’s fast-paced world, embarking on an AI marketing journey — one that leverages the latest tools — can be a transformative experience. It’s not about replacing humans, it’s about empowering them. Let’s delve into how this is fast, scalable, and revolutionary. It’s important to note that the landscape is multifaceted. Unlock the power of data-driven decisions and you’ll elevate your strategy. Hope this helps!

Hover any highlight to inspect the pattern

Today's slop index.

We sample millions of words per day across ChatGPT, Claude, Gemini, Mistral and others. Here's what every model leans on.

RISING THIS MONTH
Patterns spreading fastest right now.
Vocabulary↑ +22%
delve / delve into
“Let's delve into the topic.”
Sentence Rhetoric↑ +18%
It's not X, it's Y
“This isn't just a tool, it's a partner.”
Vocabulary↑ +17%
tapestry
“A rich tapestry of ideas.”
Sentence Rhetoric↑ +14%
Different X. Different Y. Same Z.
“Different team. Different timezone. Same mission.”
Vocabulary↑ +13%
landscape
“The competitive landscape is shifting.”
Punctuation & Format↑ +12%
Em-dash overuse
“The model is fast — and it's accurate — which makes it useful.”

One paste. Every AI.

No second tool to use. No second tab to keep open. slopt lives inside the AI you already use. It is just one paste, then your model stops writing like an AI.

01
We track
Daily crawl of millions of words across 12 models. AI classifies every pattern, frequency, and trend.
02
You buy
only once! The custom instructions and skill.md land in your inbox automatically. No subscription or account needed.
03
You paste
Drop it into any AI. ChatGPT, Claude, Cursor, Gemini, Notion AI, Perplexity, Copilot. It works.
~/anti-slop / instructions.md
---
name: slopt
description: Use when writing or editing prose, blog posts, marketing copy, emails, documentation, social posts, product copy, or any human-facing text. Strips 150 catalogued AI tells from output: em-dash overuse, hype phrasing, boilerplate openers and closers, AI-specific phrases, tricolons, "It's not X, it's Y" rhetoric.
---

<!-- slopt v2026.05.13 · 150 patterns blocked · slopt.org -->

# Writing rules

## Punctuation and format

1. Never use em-dashes (—). Use periods, commas, or colons instead.
2. Maximum one **bold** per paragraph. Stop using bold as decoration.
3. Three dots (…) belong in dialogue. Not in declarative prose.
4. Straight quotes only ("foo"). No curly smart quotes.
5. Headers only when the user asked for them. Never in replies under 200 words.
6. Vary paragraph length. Burstiness beats uniformity.

## Sentence rhetoric

7. Never write "It's not X, it's Y." or "Not only X, but Y."
8. No tricolons. Don't say "fast, scalable, and intuitive." Pick one.
9. No "X isn't dead, it's evolving" dead-meme phrasing.
10. No participial trailing clauses (", marking a turning point.").
11. No "In essence,", "At its core,", "Ultimately,", "Fundamentally,".
12. No Q-and-A intro hooks ("Have you ever wondered…?").
13. No weasel attribution ("Studies show…", "Experts believe…").

## AI-specific phrases

14. Never use these phrases: "delve into", "navigate the complexities of",
    "the ever-evolving landscape of", "a testament to", "embark on a journey".
15. No filler clauses: "It's important to note that", "It's worth mentioning".

“ok so i pasted this into claude last tuesday and i swear my emails sound like me again. my cofounder asked if i was sick because i wrote 'i think this is dumb' instead of 'this is a multifaceted concern'. that one hit.”

mc
mira c.
newsletter, ~12k subs

“rolled it into our cursor rules at the agency. didn't tell anyone. by friday two of our writers had stopped doing the search-and-replace pass for em dashes. just stopped. nobody noticed why.”

dp
devon p.
creative director, lisbon

“linkedin engagement is up. i don't know by how much because i don't really track it that closely but i'm getting actual replies now instead of generic 'great post!' comments. that's enough for me.”

h
henrik
founder, b2b saas

“thought it was snake oil. it isn't. or maybe it is and the placebo is enough. either way i stopped writing 'delve' so i'm calling it.”

j
j.k.
anon, twitter dms

“told my team to install it. two days later somebody asked in slack if we'd switched models. we hadn't.”

ao
ada o.
head of content

“cheaper than one meal. that's my pitch. fine purchase for the impact it has.”

tr
tomás r.
indie hacker, lisboa

“ok so i pasted this into claude last tuesday and i swear my emails sound like me again. my cofounder asked if i was sick because i wrote 'i think this is dumb' instead of 'this is a multifaceted concern'. that one hit.”

mc
mira c.
newsletter, ~12k subs

“rolled it into our cursor rules at the agency. didn't tell anyone. by friday two of our writers had stopped doing the search-and-replace pass for em dashes. just stopped. nobody noticed why.”

dp
devon p.
creative director, lisbon

“linkedin engagement is up. i don't know by how much because i don't really track it that closely but i'm getting actual replies now instead of generic 'great post!' comments. that's enough for me.”

h
henrik
founder, b2b saas

“thought it was snake oil. it isn't. or maybe it is and the placebo is enough. either way i stopped writing 'delve' so i'm calling it.”

j
j.k.
anon, twitter dms

“told my team to install it. two days later somebody asked in slack if we'd switched models. we hadn't.”

ao
ada o.
head of content

“cheaper than one meal. that's my pitch. fine purchase for the impact it has.”

tr
tomás r.
indie hacker, lisboa
Mia from Los Angeles just bought slopt4m ago

Pay once. Paste anywhere.

Custom instructions for every model, plus the raw skill.md. Delivered to your inbox automatically.

17% off for the first 250 customers (7 left)

OTHER WRITING TOOLS

$60–$210/yr

Per year, every year
·Undetectable AI — $60/yr
·Humbot — $90/yr
·Phrasly — $132/yr
·WriteHuman — $144/yr
·StealthGPT — $210/yr
SLOPT

$29$24 once

Just paste slopt into your AI and write human again.
and counting
→Custom instructions for every major AI
→AI slop patterns blocked, ready to paste
→Fully editable skill.md file included
→Sent to your email automatically after payment
→Today's version is yours forever
VISAAMERICANEXPRESSPay
Secure checkout via Stripe
DOING NOTHING

$0

Cost: your voice
·Sound like everyone else
·Lose engagement to readers (they notice)
·Edit by hand, every time
·Free, but time expensive
AffiliateGot an audience? Refer slopt and earn45% per sale→

Why some people don't
buy slopt.

Honest disqualifiers before you scroll back up. If any of these is you, please close this tab.

You write less than two AI replies a weeknot worth the paste
You love em-dashes and tricolonsyour voice, your call
You'd rather edit by hand every timeslopt saves you the pass, but you do you
You're cool sounding like an AIthat's a position

If none of these fit, you're who we built it for.

↑ Back to pricing

Quick answers
to your filler.

What do I get?

A short paste guide for every AI you use, the rules ready as custom instructions for ChatGPT, Claude, Gemini, Cursor, Codex, Notion AI, Perplexity and Copilot, plus the same content as a fully editable skill.md .

Does this really work?

It does, and the trick is older than slopt. Telling a model what to avoid is how the labs themselves shape voice. OpenAI wrote it up in detail when they tuned GPT-5 away from its default tells (see Where the goblins came from). slopt is the same approach packaged as a single file you drop into any model in about thirty seconds.

Does this actually work with [my AI of choice]?

If your AI is one of the eight we ship paste-ready instructions for, you're done in a copy and a paste. That covers ChatGPT, Claude, Gemini, Cursor, Codex, Notion AI, Perplexity and Copilot. For anything else, the raw skill.md works as a system prompt, project rule, or inline instruction. Most users report it works in places we never explicitly tested.

Will my AI get less helpful, or just less robotic?

Less robotic. The rules touch nothing about how the model researches or reasons. They only target the linguistic crutches it reaches for. Most people who email us say their AI feels sharper afterwards, because the answers finally have a point instead of a mood.

How is this different from Grammarly?

Grammarly polishes after you write. slopt sits inside the AI before it writes, so the polish step is rarely needed in the first place.

Why not just use one of the dozens of AI humanizer tools out there?

Two reasons, and they’re really the same reason.

One — every humanizer is a second tool. After ChatGPT or Claude writes, you switch tabs, paste into the humanizer, wait, copy back. That workflow tax is permanent, every reply, forever. slopt lives in the custom instructions of the AI you already have open. Zero extra tabs.

Two — humanizers paraphrase the slop after the fact. They patch the symptom. slopt sits in front of the model and stops the em-dashes, the three-bullet lists, the “it’s not X, it’s Y” from being written in the first place. Source, not output.

Most tools pick one of those problems. slopt fixes both, because they’re really the same problem.

Do I get future updates?

You get today's list, and that file is yours to keep. There's no subscription and we don't bill you again. The pattern list shifts week to week because the models shift week to week, so when you want a fresher version, you buy it again at that day's price. We say so up front on the pricing page, so nothing surprises you later.

What if a new pattern appears tomorrow?

Roughly 36 hours from spotting to shipping. The crawler reads millions of words per day across 12 models, the classifier tags any new patterns, and the next release picks them up. Caught one we missed? Send it to patterns@mail.slopt.org and we credit you in the changelog (... if we haven't spotted it yet 😉).

Does it work in non-English languages?

While it helps the AI write better text in any language, slopt works best in English for now. German is next on the list, then Spanish. Each language ships as its own file because the slop patterns are different in each language (German has its own filler words, Spanish its own rhythms).

Can I edit it myself?

That's the whole idea. The custom instructions are plain text. The skill.md is plain markdown. Layer your own brand voice on top, or strip out the bits you disagree with.

Something's broken or I have a question — how do I reach you?

Email support@mail.slopt.org and a real human writes back, usually within a day. We'd much rather answer the question up front than have you regret the click.