About — How the Voice Score Works and Our Integrity Note

What Humanize Like Me is

Humanize Like Me rewrites AI-generated text so it sounds like you. You build a visible, editable voice card from your own writing, and every rewrite follows it. You review changes sentence by sentence.

It's for people who use AI to draft and want the result to read like their own work, where that's allowed.

How the “sounds like you” score is computed

The likeness score runs from 0 to 100. It's rule-based. There's no hidden model deciding it, and each score comes with plain-English reasons. It compares the rewritten text with your voice card and samples on these parts:

Sentence length compared with yours.
Rhythm: how much your sentence lengths vary.
Contractions: how often you use them.
Word level: how plain or formal your vocabulary is.
Punctuation habits (em dashes, semicolons, parentheses, exclamation marks) compared with your samples.
Small-word habits such as and, but and so.
Never-words: anything on your “never use” list that slipped in.

What the score is not

It isn't a measure of whether text will pass any AI detector, and it can't prove who wrote something. It tells you how close the text is to your own writing, which is the thing readers notice.

Integrity

Humanize Like Me is for making your own drafts sound like you: emails, posts, reports, and essays where your school or employer allows AI help. If your class or workplace bans AI-assisted writing, this tool doesn't change that rule, and you shouldn't use it to hide AI use.

What we don't do

We don't promise to bypass or pass Turnitin, GPTZero, Originality or any other detector.
We don't claim text is “undetectable” or “100% human”.
We don't help submit work that isn't yours.
We don't sell your text or use free rewrites to build a profile of you.
The free tone checker is a writing check. It isn't a detector.

Methodology honesty

Rewrites come from an AI model, and models make mistakes. The fact lock catches numbers, dates, names, quotes and URLs, but it can't catch every kind of error. Read what you publish. Comparisons on our site are dated and based on what we could verify; where we couldn't, we say so.