Unofficial project — not affiliated with blader/humanizer or any AI vendor.
Runs 100% in your browser — nothing is uploaded

Your writing has a tell. Find it, then fix it.

Blader Humanizer scans text for the patterns language models can’t stop reaching for — em-dashes, the rule of three, “delve” — scores how AI it reads, and rewrites it to sound like a person. Free and open.

draft.txt
0 words 0 chars
0%
Pattern Score
Paste text and detect to score it.

Detected patterns

0 patterns
How it works

Three steps. No account. No upload.

Everything runs locally in the page. Your draft never leaves the browser tab.

01

Paste

Drop in anything an AI might have written — a cold email, a blog intro, a PR. The sample loads pre-filled so you can try it instantly.

02

Detect

Eleven heuristics flag the tells — em-dash density, triads, hedged meta-commentary, hype vocabulary — and roll them into one 0–100 Pattern Score.

03

Rewrite

Hit Humanize to strip the dashes, swap the buzzwords, break the triads — then watch the score drop. Undo anytime; you stay in control.

AI writing patterns

The fingerprints models leave behind

A living library of the constructions that quietly scream “generated.” Here are six of the loudest.

Punctuationvery common

Em-dash overuse

Models pepper sentences with — long dashes — where a person would use a comma or just stop. Density is the giveaway.

It’s fast — really fast — and simple.
It’s fast, really fast, and simple.
Structurevery common

The rule of three

A compulsive reach for triads: “simple, powerful, and elegant.” One is rhetoric; a paragraph of them is a tell.

Fast, simple, and reliable.
Fast and reliable.
Dictionnotorious

“Delve” & friends

Delve, tapestry, realm, leverage, robust, navigate, underscore. A vocabulary far more common in models than in people.

Let’s delve into this rich tapestry.
Let’s look at this.
Fillercommon

Hedged meta-commentary

“It’s worth noting,” “it’s important to note” — throat-clearing that announces a point instead of making it.

It’s worth noting that speed matters.
Speed matters.
Structurecommon

“It’s not just X, it’s Y”

Negation-then-elevation. The cadence sounds profound and means almost nothing — a signature flourish.

It’s not just a tool — it’s a movement.
It’s a useful tool.
Fillercommon

Grand openers

“In today’s fast-paced world…”, “In the realm of…”. Scene-setting that delays the actual sentence.

In today’s fast-paced world, speed wins.
Speed wins.
FAQ

Questions, answered

No. Blader Humanizer is an independent, unofficial web tool inspired by a popular open-source Claude Code skill. It isn’t affiliated with, endorsed by, or connected to blader/humanizer, Anthropic, OpenAI, or Google.
Never. Detection, scoring, rewriting and the ZIP export all run in JavaScript inside your browser tab. There’s no server, no account, and no analytics on your content.
It’s a transparent heuristic, not a black-box classifier. Eleven detectors each contribute a weighted amount based on how many times their pattern appears, lightly normalized for length. A high score means many recognizable tells — it’s a writing-quality signal, not a verdict on authorship.
That’s not the goal. The tool is built to make writing clearer and more human, not to game any specific detector. Use it to edit honestly — strip the filler, vary the rhythm, cut the buzzwords.
It builds a small, ready-to-use skill folder (SKILL.md + README) on the spot and downloads it as a ZIP — entirely in your browser. Drop it into Claude Code so the model applies the same humanizing rules whenever you draft.
Yes — free and open. No sign-up, no paywall, no usage cap.
Install

Put the humanizer in your workflow

Use it here, or generate the Claude skill and let it clean every draft you write.

Blader Humanizer — detect & remove AI writing patterns