Humanize ChatGPT text

ChatGPT drafts are the easiest to spot and, in our tests, the easiest to fix. Here is what gives them away and how to rewrite them.

Open the humanizer. It is free: try 400 words without an account, or sign in with Google for 6,000 words a day, up to 1,200 words per paste.

How ChatGPT writes

Every chatbot has habits. These are the ones AI detectors and readers pick up in ChatGPT drafts:

Openers like "In today's rapidly evolving world" and closers like "In conclusion".
"Delve", "crucial", "it's important to note", "furthermore" and "moreover".
Em dashes and lists of three in almost every paragraph.
Paragraphs of the same length, each ending on a summary line.

From our own GPTZero tests in October 2026. These are small tests, so treat them as a guide, not a promise:

A 300-word ChatGPT essay on social media and mental health: 100% AI before, 0% after, in two separate runs.
The introduction of a ChatGPT literature essay: 7% and 0% AI after two runs.
Across six runs on ChatGPT drafts, results ranged from 0% to 14% AI.

Copy the ChatGPT answer without the chat around it ("Sure! Here's...", "Let me know if...").
Open the humanizer, pick the style that fits, and paste up to 1,200 words.
Penned rewrites each paragraph several times, scores every version with an AI detector and fact-checks the best one.
Delete any sentence that only restates the one before it. ChatGPT pads, and padding reads as machine-written.
Check the result in the AI detector.

Often, from its habits as much as from detectors. A humanizer removes the habits; adding your own examples and opinions does the rest.

Yes. Penned works on the text, not on which model wrote it. We add new chatbot drafts to our test set as models change.