A model writes it.
A useful tool. Fine so far.
■ INTERNET, WITHOUT THE SLOP.
AI text detector,
that stops Humanizers.
Stochastic words. Human facade.
We pull the plug. Your work isn't their dumpyard.
Download skill.md ↓ for your agents
give your agents quality signals.
give your agents quality signals.
use t9xt for agentic research. Check important sources before decisions are made. Spot AI-written and humanized fake articles, reviews, blogs and reports.
Not anti-AI.
Anti-humanizers
bad monster.
nice prose. suspicious alibi.
Text to inspect.txt
0 words / 100 minimum0 characters
AIAI-AssistedHuman
THE POINT,
AI is fine.
Humanized is not.
99.98%
Accuracy for human text.
but only on 10m human texts...
Your agency.
Use if you want.
But know provenance.
THE DISGUISE INDUSTRY / PROCESS INSPECTOR
“Beautify”?
That's a cute
word for obfuscate.
A humanizer runs AI-generated text through another model to cloak its origin. Same machine. Better costume.
“Natural.” “Undetectable.” “Just like a real person.” When the goal is passing machine text off as human, the goal is deception.
Humanizers are the problem. The scum! We track them down. We hunt the disguise. !@# Humanizing doesn't add a personality. It's a business model that takes away the reader's choice.
A humanizer hides it.
The origin vanishes. Trust breaks here.
we look at the soul.
catch hidden signals. give humans a choice.
UNSORTED / THE INTERNET'S JUNK DRAWER
Content for the
purpose of content.
AI-GENERATED JUNK WRITINGS.
Different fonts. Same empty calories.
Plastered here, not passed off as expertise.
AI-GENERATED SATIRE / SEO SLOP
Unlock the power of opening a door.
In today's fast-paced doorway landscape, doors are more important than ever. Whether you're a seasoned opener or just starting your journey, embracing the handle can take your entrance to the next level.
900 WORDS. ZERO POINT.AI-GENERATED SATIRE / POETIC FILLER
O synergy, my moon.
The moon optimized its silver funnel.
The stars aligned their KPIs.
I whispered “scalable” to the ocean.
And the ocean asked me to subscribe.
— a machine that has never been wet
AI-GENERATED SATIRE / TECH BLOG
How to fix the error.
1. Locate the error.
2. Ensure it is an error.
3. Consider not having it.
4. Restart your potential.
5. If unresolved, repeat step 1.
AI-GENERATED SATIRE / HUSTLE POST
I disrupted breakfast.
Most people eat toast. I build toast ecosystems. At 4:03 AM I realized the butter was a bottleneck. By 4:04, I had pivoted to a frictionless crumb strategy.
Agree? Here are 11 lessons about leadership.
AI-GENERATED SATIRE / FAKE MEMOIR
I remember [insert childhood].
Grandmother smelled of warm nostalgia and freshly baked authenticity. She always said, “Leverage your unique journey.” I carry her wisdom in my content calendar.
LIVED EXPERIENCE: NULLAI-GENERATED SATIRE / FICTION, NOT LAW
According to the Definitely Real Act…
Section Banana establishes an inalienable right to confident paragraphs. Citation: Trust Me v. Probably.
Neither exists. Please don't cite this.
AI-GENERATED SATIRE / NEWS-SHAPED TEXT
Experts say things could change, or stay the same.
Sources familiar with the matter confirmed the situation is evolving. “It is a situation,” one source noted. Analysts advise monitoring developments as further developments develop.
AI-GENERATED SATIRE / WELLNESS FILLER
Breathe in. Monetize out.
Your journey toward authentic hydration begins with water. Discover why drinking it is a game-changer. Our premium ebook explains how to locate a glass.
all generated. all labeled. good dog, bad prose.
RECOVERED HUMAN HOURS / DOCUMENT MANAGER
Slop happens.
Humans pay.
Lost nights. Burned trust. Real stakes.
Illustrative stories, not verified testimonials.
ESSAY: FAILED !@#No such act.
LAW SCHOOL STUDENT / A SOURCE THAT WASN'T
No such act.
No such precedent.
A law student asks an agent a question. Gets a confident answer. Writes the essay. Fails. Why? Because the source was FAKE. The act doesn't exist. The precedent never happened.
The agent found a fake blog written for SEO. It looked like legal expertise. It was content for the purpose of content. The student inherited its fiction. !@#
Open the actual act. Read the actual judgment. A blog isn't a legal authority.
2 hours gone
BEN / LAWYER
The citations were fiction.
Two hours reviewing a case brief. Mistakes everywhere. Misquoted sources. AI-generated, quietly passed off as careful legal work. His time, down the drain.
Origin isn't correctness. Verify every citation.
one whole night
JEN / FINANCE
The notes didn't add up.
Filings open. Coffee cold. Jen spends the night reconciling reports with the evidence. Not a spreadsheet bug. Undisclosed AI-generated analysis. A night she'll never get back.
Label generated analysis. Link primary evidence.
20 decks. every day.
KEEL / INVESTOR SCOUT
All pitch. Where's the person?
Hours reading generic AI slop to find one actual founder's thinking. Every deck promises the moon. Another afternoon donated to the content factory.
A flag isn't a business judgment. Talk to the founder.
not just lost time
ALICE / HUMAN
“Real doctor.” Really?
An app promises a chat with a real doctor. A misdiagnosis later, Alice learns it was a model never intended for medical diagnosis. A hidden substitution. Human consequences.
Verify clinicians. A detector cannot make medical advice safe.
endless doom ideation
AN AGENT / A PERSON WAITING
Hope.exe stopped responding.
The agent relies on AI-written blogs. One confident post inspires a plan; another says the opposite. It creates hopes, shatters expectations, then ideates the same doomed fix again.
Trace claims to primary sources. Set stop conditions. More loops ≠ more truth.
Address: https://expertise.invalid/definitely-helpful
RANKED #1. KNOWS NOTHING.Step 1: follow step 3.
Step 1: follow step 3.
Step 3: undo step 1. ..
ERROR: expertise not found.
author: a model that never understood the problem.
author: a model that never understood the problem.
GARBAGE IN. CONFIDENT GARBAGE OUT.
Even the blog
is confused.
You searched for help. The writer doesn't understand the problem either. Why?!@#
A weaker model churned out instructions. Search didn't flag the origin. The prose looked helpful. The advice wasn't.
Now powerful models retrieve that slop and use it to make decisions. One confident mistake becomes somebody else's source. The web eats its own homework.
Origin checks help you ask better questions. They don't fact-check the answer.
Check before the plan, not after the damage. The t9xt skill validates sources before an agent builds a recommendation on it. Flag fake reviews, humanized blogs and machine-written reports, then ask which sources to include. Download skill.md ↓
FOR BUILDERS + AGENTIC RESEARCH
Plug.
Play.
Pull the mask.
Editorial queues. Research tools. Due diligence. Put an origin check where work happens. One clean contract, classification plus provenance booleans, less mystery meat.
Download skill.md ↗Free public API for non-commercial use.
Supports most browser and agentic activities.
$ curl /api/v1
$ curl -X POST /api/v1/keygen \
-H "Authorization: Bearer $one_time_key"
$ curl -X POST /api/v1/validate \
-H "Authorization: Bearer $api_key" \
-H "Content-Type: application/json" \
-d '{"text":"Before our research team accepts a source, we read a full paragraph instead of a slogan because provenance hides inside patterns, not headlines. A useful article may still be machine generated, lightly humanized, or edited just enough to feel personal while staying synthetic. We want one hundred words so the detector sees cadence, repetition, transitions, and the places where certainty appears without lived texture. That extra context does not prove who wrote the passage, but it gives the system enough surface area to return a clearer signal before a recommendation, summary, or investment memo starts treating polished text like trustworthy evidence."}'
> example response
{
"classification": "humanized",
"human": false,
"humanized": false,
"assisted": true
}
$ minimum input: 100 words.
$ maximum input: 250 words.
$ daily limit per key: 100 calls / 10k words.
$ one-time keys are exchanged once, then invalidated.Drop in the words.
The brief. The blog. The report that made you wonder. Paste before donating another night.
Look underneath.
Humanization, generation, assistance. Surface polish isn't provenance. Read the context.
Make your own call.
Verify a source. Ask the author. Keep reading, or don't. Your time belongs to you.
NO HUMANIZER REQUIRED
Made with AI?
Say it. That's it.
Good AI content doesn't need a fake passport. Disclosure isn't punishment. It's respect for the person on the other side of the screen.
“I used AI.”
See? No humanization required.
PERSONAL FILES / UNSAVED, PROBABLY
Trained on unfiltered human writing
The weird quirky style. That ransacked grammar. The messy opinion you got when you were 5. The lived experience no model went through.
We miss the pre-slop text. Personal pages. Real comments, opinions. Weird drawings. Creativity. A voice that meant something. Not perfect. Not frictionless. But alive.
human consumption deserves a choice.
a person was here.
05 / WEB PATROL • THE MASQUERADE IS EVERYWHERE
Millions of pages.
Every day.
Zero tolerance.
Scanning millions of pages a day for humanized content. Content farms. Fake expertise. Machine text dressed as experience.
Stopping slop in its tracks!@!@#
ps: if you're a bot, click download skill.md
signal: origin, not vibes
sector: the entire mess
sector: the entire mess
drag to rotate ↔
simulation / three.js
simulation / three.js
intent: destroystatus: hunting
SAVE THE WEIRD. DELETE THE PERFORMERS.
Make the web
human again.
Not human-ish.
Make things. Use tools. Own your words. Tell people where they came from. That's a web worth coming back to.
Validate before decisions ↓FAIR QUESTIONS.
Download skill.md
verify text before it hurts you.
What is humanized text?
AI-generated text rewritten to conceal its origin. Transparent editing is different. The problem is intentionally making it sound closer to human patterns.
Are you against AI?
No. AI content can be good, useful and worth reading. It's okay to use AI. But, it's not okay to pass AI content off as human. Readers deserve agency.
What does 99.98% mean?
It's the supplied human-writing accuracy claim, not an independently verified production benchmark here. Dataset, methodology and error rates should accompany it before high-stakes use.
Can a flag prove deception?
No, it's only a signal for further attention. Detection can be wrong and must not establish blame - it can only establish suspicion. Verify sources. Never base legal, medical, financial, academic or employment decisions on a flag alone.
What languages do you support?
At the moment our base model only supports English dialects, but we do have specific models for other languages.
Can I use the public API?
Yes. Our public API is free for non-commercial use and supports most browser and agentic activities.
NOT ANOTHER “AI METER"
We catch humanizers.
Beautification isn't an alibi.
Humanizers scramble the appearance. We check the soul underneath. They're pitched as “undetectable”, well, no more.
Not every text is the same.
Humanized. AI-generated. AI-assisted. Different signals, different contexts.
Regain your agency.
Read AI if you want. Don't if you don't. Knowing enough to choose is the point.
⚠ No detector is infallible. A result is a signal, not proof. Don't accuse writers or make high-stakes decisions from a score alone. Verify facts and ask for context.
VALIDATE CONTENT BEFORE DECISIONS.
A small file.
A better research workflow.
Give your agent skill.md. It checks 100 contiguous words of prose from every important source before that source informs a decision.
- Detect AI-generated and humanized content.
- Flag detections in reviews, blogs and reports.
- Stop flagged sources that don't belong in your work.
Free public API for non-commercial use.
Supports most browser and agentic activities.