Here's how to make ChatGPT write like a human and beat AI detectors

In this post, we're going to see how you can make ChatGPT, Bard, and any other LLM write like a human and bypass most (if not all) AI-generated content detectors for free.

In a previous post, I reviewed some of the best AI text detector tools out there. Today, however, we're going to look at how we can make ChatGPT, Bard, and other models write text like a human and which tools allow us to tweak the generated output so that it easily bypasses common AI detector tools.

🤝
Disclosure: I earn a commission if you make a purchase through the links in this post, at no extra cost to you. Please note that this is not a sponsored post, so the views expressed are entirely my own.

TL;DR

No time to read but want to "humanize" your AI generated content? Here's a great tool called Undetectable that you should try:

Undetectable.AI - Content Humanizer

Undetectable AI humanizes your content to match the quality of human writing.

Try Undetectable.AI Now
💡
But hey, come back and go through the content of this post as I explain some key concepts about how AI content humanizers bypass AI detectors.

Introduction

While some people want to hide the fact that their content was generated with artificial intelligence LLMs such as ChatGPT and other models, other people want to detect if the content was actually written by a human or generated.

You could think of a college professor trying to determine whether their students submitted generated essays or they actually made the effort to write them themselves. Or a client who's paying copywriters (content writers) for supposedly original work.

Whichever the case, today there are two forces fighting for each side. There are tools that try to hide AI-generated content and others that try to catch it.

Perplexity and burstiness

To understand how AI text detectors work, we need to quickly go over the terms Perplexity and Burstiness.

Perplexity determines how unpredictable a given text is. It comes from the word perplex which means confuse, baffle, etc. AI-generated text produces "low perplexity" content, making the output more easily readable.

Burstiness refers to word variation and sentence structure in a given text. The AI-generated content has "low burstiness" because the way LLMs generate sentences is by adding the next most likely word given all previous words making the text easier to read but lacking character or identity.

💡
AI-generated text is easy to read because it has low perplexity and low burstiness.

AI detectors among other techniques look for low perplexity and low burstiness to detect if the text is AI-generated. We're going to now look at a popular tool that also works to increase perplexity and burstiness while still retaining a logical structure in a given text.

Humanizing AI content

Technically, we're not going to make ChatGPT write like a human, but instead, we're going to take its generated text and transform it in a way that is more human-like. A trending keyword today is Humanizing AI-generated text, which means exactly what we just explained.

One such tool that will help us accomplish this task is Undetectable.ai which promises to turn AI-generated text indistinguishable from human text.

We turn your flagged AI content into high-quality writing, that bypasses AI detectors and is indistinguishable from human-written text. - Undetectable.ai

For this example, I am going to ask ChatGPT to write a short paragraph about the beautiful city of Montréal, Quebec:

Montreal, a vibrant city in Canada's Quebec province, is a captivating blend of European charm and North American energy. With its rich history, diverse culture, and bilingual character, Montreal offers a unique and welcoming atmosphere. The city is renowned for its stunning architecture, world-class dining scene, and a dynamic arts and music community. From the cobbled streets of Old Montreal to the bustling neighborhoods, Montreal's distinctive fusion of tradition and modernity makes it a captivating destination for locals and visitors alike. - ChatGPT

After using the above text, all AI-detector tools came back to detect at least 51% or more AI-generated words and sentences.

I am now going to copy and paste the paragraph above into Undetectable.ai's tool and then click on the 'Humanize' button as shown below:

Here's the 'humanized' output:

Montreal, a city located in the province of Quebec, Canada is a combination of European allure and North American vitality. With its background diverse cultural landscape and bilingual nature Montreal boasts a warm and distinctive ambiance. The city is renowned for its structures top notch culinary scene and thriving arts and music community. From the cobblestone streets of Old Montreal, to the neighborhoods that buzz with activity Montreals fusion of tradition and contemporary elements creates an alluring appeal that captivates both locals and tourists alike.

Result? It works! AI text detector tools indicate that the text is mostly human-written.

Undetectable.AI

Undetectable.AI - Content Humanizer

Undetectable AI humanizes your content to match the quality of human writing.

Try Undetectable.AI Now

Go ahead, try it yourself now. You'll just need to enter and verify your email before you can start generating 'humanized' versions of ChatGPT and other LLM outputs.