AI Text Humanizer Pre-Checker

Check Your Draft's Sentence Rhythm & Repetition • Not an AI Detector

This is not an AI detector and can't tell you who or what wrote a text. Research on "perplexity and burstiness" explicitly finds these signals unreliable for that purpose — humans often write with low variation too, especially in formal contexts. What this tool can do: flag stylistic patterns — uniform sentence length, repeated words, overused AI-style vocabulary — that often make a draft read as mechanical, so you can edit it to sound more like you. A single flagged word proves nothing; density and pattern across the whole draft is what matters.

Your Draft

0 characters

0 words

0 sentences

Sentence Rhythm

A mix of short and long sentences generally reads as more natural. Very uniform lengths can feel monotonous — this is sometimes called "burstiness" in writing research.

Average Length

Length Variation (Std. Dev.)

Assessment

Sentence Length by Position

Each bar is one sentence, height = word count. Bars of very similar height throughout suggest a uniform, potentially monotonous rhythm.

Repeated Words

Content words appearing often relative to the draft's length. Some repetition is normal — look for words that feel like a verbal tic once you notice them.

Repetitive Sentence Openers

Starting many sentences the same way is a common source of monotony.

About

What This Isn't

Contact

About the Pre-Checker

This tool analyzes a draft for stylistic patterns that tend to make writing feel mechanical or monotonous: uniform sentence length, repeated content words, repetitive sentence openers, and vocabulary that's documented as disproportionately common in LLM output. It's meant for self-editing — a way to spot-check your own AI-assisted draft before you revise it into something that sounds more like you, not a way to judge someone else's writing. All the analysis (sentence splitting, word counting, pattern matching) runs directly in your browser using plain text statistics — nothing you paste is ever uploaded anywhere.

What This Isn't

Not a detectorThis tool cannot determine whether a text was written by a human or an AI, and doesn't attempt to. It measures surface-level writing patterns, nothing more.

Why notPublished research on AI-text-detection signals like burstiness explicitly notes they're unreliable for that purpose — formal human writing often shows the same low variation attributed to AI, and detection accuracy degrades further as models improve.

One word proves nothingA single "delve" or "underscore" in your draft means essentially nothing on its own — these are ordinary English words. Density and clustering across a whole draft is a more meaningful signal than any single occurrence.

The vocabulary list changesWhich words get flagged as "AI-sounding" shifts with every model generation — this list reflects patterns documented as of when this tool was last updated, not a permanent rule.

Real-world stakesPeople have faced real consequences — academic penalties, rejected work — from writing flagged on thin evidence like a single word. Use this tool to improve your own writing, not to accuse anyone else's.

Send Feedback

For any questions, bug reports, or feature requests, please drop your comments below.