How to make ChatGPT undetectable

ChatGPT text gets flagged because of its habits: the even rhythm, the stock words, the friendly wrapper around the answer, and general lines where a person would be specific. To make it undetectable, take those habits out. Paste the text into Hemmingway's free AI humanizer, or ask Hemmingway for the writing in the first place.

Hemmingway writes with Hemmingway-1.5, the undetectable AI, trained to write like a person. In our test, 90% of what it wrote passed GPTZero and four other top AI detectors on the first try, and by its Hugging Face card, 89% of its rewrites of AI text passed the same detectors.

What gets ChatGPT text flagged

• The wrapper. "Certainly! Here's a draft you can use", then the text, then "Let me know if you'd like me to adjust the tone".
• Stock words. "Delve", "crucial", "moreover", "it's worth noting", "I hope this message finds you well".
• Even rhythm. Sentences of the same length, paragraphs of the same size.
• Everything in threes. Three reasons, three adjectives, three bullet points.
• Nothing specific. "At your earliest convenience" where a person would say "by Friday".
• The wrong register. A two-line message to a friend that reads like a cover letter.

AI detectors read exactly these patterns. In our test, GPT-6 Astra, the ChatGPT model we tested, passed GPTZero and four other detectors together 20% of the time, on 80 everyday asks.

The fix, step by step

With the AI humanizer

1. Copy ChatGPT's answer. Leave out its opening and closing lines, and keep only the text you need.
2. Paste it into the AI humanizer. It rewrites it right on the page, with no account.
3. Say who will read it and how it should sound. "My landlord, short and polite" is enough.
4. Add one real detail. The day, the name, the number that only you know.
5. Use it. The rewrite keeps everything the text said, in a person's voice.

Or write it with Hemmingway instead

Ask Hemmingway the same thing you would ask ChatGPT. Hemmingway-1.5 gives you the text and nothing around it, in the register the reader needs, with no rewrite to do afterwards. The free tools start you off: the AI essay writer for a whole essay, the AI email writer for email, and all the writing tools for the rest. The Hemmingway app also searches the web, reads the pages and cites its sources.

Or fix it by hand

Cut the opening and closing lines, put in the specific facts, vary the sentence length and drop the stock words. How to make AI text sound human goes through each habit.

An example

A typical ChatGPT-style answer to "text my landlord that the kitchen tap is leaking and ask when someone can fix it" looks like this:

Certainly! Here's a polite message you can send to your landlord:
Dear [Landlord's Name], I hope this message finds you well. I wanted to bring to your attention that the kitchen tap has been leaking. I would greatly appreciate it if you could let me know when someone might be available to address this issue at your earliest convenience. Thank you for your time and assistance.
Let me know if you'd like me to adjust the tone!

Through the AI humanizer, for a landlord, short and polite, a rewrite along these lines is what to expect:

Hi, the kitchen tap is leaking. When could someone come and fix it?

Same request. The wrapper, the stock opening and the padding are gone, and it reads like a text a person sends.

Why the writer matters

Detectors catch most of what the big models write. On the same 80 everyday asks, run once each through GPTZero, Fakespot, Vanguard, WasItAIGenerated and MELD:

Training is what changes the result. Hemmingway-1, our first model, passed 7% of the same detector test. Hemmingway-1.5, built on it at the same size, passed 90%. The method is in our AI detector test, and Hemmingway vs ChatGPT puts the two side by side.

Try HemmingwayDownload the app

Common questions

Can ChatGPT text be made undetectable?

Yes. Rewrite it so its habits are gone, with the AI humanizer or by hand. By its Hugging Face card, 89% of Hemmingway-1.5's rewrites of AI text passed GPTZero and four other top AI detectors.

Does GPTZero detect ChatGPT?

Its site says it flags text from ChatGPT, Gemini and Claude. In our test, GPT-6 Astra, the ChatGPT model we tested, passed GPTZero and four other detectors together only 20% of the time. How to bypass GPTZero covers what it measures.

What is the quickest way to make ChatGPT undetectable?

Paste the text into the free AI humanizer. It rewrites it on the page with no account and no card, and keeps what the text says.

Is there an AI that is undetectable from the start?

Yes: Hemmingway-1.5. In our test, 90% of its writing passed GPTZero and four other top AI detectors on the first try, more often than the 86% that real human writing got.

Is what I paste used for training?

No. Chats, including anything you paste in, are never used for training.