ChatGPT is quick at writing content, which makes it an excellent study partner. And as educational institutes embrace AI, many students now use these tools for writing assistance, such as refining a few sentences, generating title ideas, and preparing outlines.
However, a problem many students face with these tools is their robotic language which leads to AI detection in content. And although students shouldn’t use ChatGPT to write what they’re supposed to write themselves, AI detection can still make its way into the content even if they only use it for writing assistance or sometimes when they don’t use it at all.
To avoid this detection, students turn to humanizing the content, which is basically rewriting the AI content to make it sound and feel more human-written to read, allowing the content to bypass AI detection. However, humanizing AI content is quite challenging. Because it involves a lot of AI writing patterns, students don’t exactly know how to make it sound human-written and break these patterns.
If you’re also struggling to humanize AI content, don’t worry. In this article, I will show you the exact techniques for humanizing AI content to bypass detection as well as make it sound natural. I will also share the easiest way to humanize AI content. So, keep reading to find out.
What is AI Detection?
Understanding what AI detection is and how it works is important, because a lot of people get it wrong. They think that an AI detection score measures how “much” of the text is AI-generated. For example, 20% detection score in a text of total ten sentences would roughly mean two sentences are AI-generated. But that’s not how AI detection works.
Instead, AI detection is a measure of how “likely” it is that your content was written by an LLM (large language model) like ChatGPT. The more the detection score, the greater the “chances” of it being AI-generated—not its “extent.” For example, a 20% score in a text of ten sentences means, the tool is 20% convinced that the text is AI, and 80% convinced that it’s not AI. Here’s a breakdown:
0% means the text is very unlikely to be AI
50% means the tool is uncertain about detection
100% means the text is highly likely to be AI
AI detectors also highlight some parts of the text. This highlight usually means that the flagged parts are suspicious, resembling AI writing, but aren’t necessarily AI-generated.
How to Get Rid of AI Detection?
Getting rid of AI detection involves humanizing your text to make it less robotic. The humanizing process involves rewriting the AI text using humanizing techniques to remove the AI writing patterns present in it.
There are three different ways to humanize AI text, including:
Using a text humanizer like HumanizeAI.net. This is the easiest and most effortless method.
Using prompt engineering. This is easy but the least effective way to humanize content.
Humanizing manually. This is challenging but the most effective way to humanize.
Let’s take a closer look at each of these techniques.
1. Humanizing Using HumanizeAI.net
The easiest way to humanize AI content is by using a text humanizer tool like HumanizeAI.net. It rewrites your ChatGPT content and removes its AI patterns that make it robotic, giving you a fully humanized text free of AI patterns.
HumanizeAI.net works much like how you would manually humanize AI content, except that it's automatic, of course. It uses similar humanizing techniques to manual humanizing, such as removing common AI words and phrases as well as varying sentence structure. You just input your text and the tool humanizes it for you.
Below, I explain how to use it.
How to Use HumanizeAI.net
Using HumanizeAI.net only requires a couple of steps, as follows:
Paste your text: Copy and paste your text into the tool or upload the file containing the text using the “File Upload” button.
Click the Button: Click the “Humanize Text” button at the bottom.
Copy Output: After a while, the tool will present the humanized text in the output field, which is below the input field. Copy this text or download it as a file using the “Download Text” button.
HumanizeAI.net also gives an AI-detection score below the output field after humanizing the text. This saves you from the hassle of checking for detection in a separate detector every time you humanize a text. Check the score to confirm the text is human-written.
2. Humanizing Using Prompt Engineering
There’s another way to humanize ChatGPT text, and that is by using ChatGPT itself. To do this, you need to engineer prompts to instruct the tool to write in a certain way. However, it’s not the most effective way to humanize content, especially if you’re giving the tool a piece of text instead of asking it to generate a newer one.
In prompt engineering, you specify all the details of the content you want and ask the tool to write in a specific style instead of its robotic style. The goal of the prompt is to force the tool to break free of the robotic writing style, so you need to advise against common AI patterns.
Furthermore, prompts that you engineer can be lengthy, but the more detailed it is the better the results will likely be. You can divide your prompt into the following parts, as shown in the example prompt below:
Example ChatGPT Prompt:
First, clarify what you want the tool to do and what the prompt is for. For example, “I want you to write an article according to the instructions given below.” This gives the tool the necessary context to follow the prompt instructions more clearly. You can also explain the goal of the prompt briefly to make things clear. “The goal of this prompt is to generate human-written content with minimum AI-detection.”
Content Requirements: This includes details about the content, such as the title and audience type.
Title: “How to Become a Better Writer”
Content Type: Blog Post/Social Media Post/etc.
Audience: Adults, content writers, professionals. Assume the audience has sufficient knowledge.
Author Persona: Author persona is the personality you want ChatGPT to assume as the writer. Should it write as a teacher? A freelancer? Or a business professional? The tone and voice also define the personality.
Personality: You are an experienced content writer
Tone: Your tone is professional, semi-formal, and humorous. Avoid an overly formal tone.
Voice: You’ll write in first-person voice. Keep the voice natural and conversational.
Goal: The goal of the article is to inform, educate, and motivate readers.
Writing Style: This is the writing style of the content you want the tool to adopt. It is the most crucial part of the prompt because this helps the tool deviate from its robotic style. Include the following aspects:
Natural Vocabulary: You must not use the repetitive and cliche words and phrases used by AI tools. Keep the vocabulary natural and also avoid overly formal expressions.
Sentence Structure: Avoid stiff sentence structure used by AI tools. Instead, use a mix of varying sentences.
Text Flow: Minimize the uses of dashes, especially em dashes. Content should feel natural to read, not dramatic.
Examples: Include relatable, real-world examples but avoid adding excessive examples.
Content Instructions: Include more instructions about the content, like headings, subheadings, or keyword additions, etc. The more the instructions, the more tailored your content will be. But these are all optionals, so you can leave them to the chatbot if you want.
Introduction: (i.e., open the article with this intro: [example introduction])
Main Points: (three-four main points you want the tool to cover, such as common mistakes, misunderstandings, etc.)
Heading Structure/Suggested Headings: [example heading structure]
Word Count: 1200-1500
FAQs: Write 3-4 FAQs at the end. [Or add example FAQs]
Things to Avoid: This part of the prompt tells the tool about what to avoid doing while writing. Instruct the tool to avoid specific robotic writing patterns that cause AI detection.
Flowery Language: Use simple language and avoid fancy words.
AI-Isms: Avoid commonly used AI words and expressions such as “tapestry.”
Em Dashes: Use alternative punctuation.
Furthermore, consider adding examples in your instructions, like in the AI-isms part. Large language models (LLMs) like ChatGPT understand the command more clearly when they’re given examples.
Once you enter the command, the tool will respond accordingly. It might ask for a few clarifications before generating the content. Once done, copy the output text and check its detection status. The detection score should be lower than average AI content, ideally around 40-50%.
Drawbacks of Humanizing Using ChatGPT
A drawback of humanizing content, or generating a humanized content, is that it isn’t very effective. This is because AI tools are by nature trained to write the way they write, which leads to detection. The common AI patterns are like the mechanism on which these tools respond to you and write content. Breaking these rules and trying to write content in a completely different way can cause the tool to write clunky and disjointed text.
If you try too hard to force humanization, it might just break the text, generating nonsensical content with no coherence or cohesion, in attempts to make it unpredictable by AI detectors. In other cases, the tool might produce very little content, such as one small paragraph per section, especially if the humanization is rigorous.
Due to these limitations, you can’t have ChatGPT fully humanize a piece of text while maintaining its quality. A better approach in this case is to prompt the tool to generate mildly human-written content, such as by using the prompt given above, and then humanizing the rest manually or using a text humanizer to completely reduce the AI detection.
3. Humanizing Manually
The most effective way to humanize ChatGPT content is to do it manually. Although manual humanization is slower and can be challenging to do, it gives you more control and precision over the text and the outcome. You can rewrite the text yourself while ensuring adequate quality.
1. Avoid AI-Isms
A common sign of ChatGPT content is the presence of words and phrases commonly used by it, also known as AI-isms. These expressions contribute to the text’s detection score, making it one of the AI patterns to remove. To remove this pattern, read the AI-generated text and find and remove these repetitive expressions.
2. Vary Sentence Structure
Sentence structure refers to the construction of a sentence, such as the “subject + verb + object” structure. ChatGPT favors a very rigid and stiff writing style in which most sentences follow a similar structure. This lack of variation is due to low burstiness in chatbots’ writing that makes the content both mundane and predictable for AI detectors. Yet, varying sentence structure is one of the most effective ways to humanize ChatGPT content.
Removing this AI pattern requires you to vary sentence structures as much as you can, and there are a couple of ways to do that, including:
Changing Sentence Clause Order: Writing the second clause first in the sentence, followed by the first clause, can help break the usual robotic sentence structure.
Changing Sentence Voice: Passive voice is common in AI content. While both active and passive voices are used, ChatGPT writes a fair amount of sentences in passive voice. If you prioritize active voice sentences in your content, it can help vary sentence structure.
Tweaking the Sentence Structure: Make deliberate changes to sentence structure and try to break free of the standard syntax.
3. Increase Functional Words Density
Functional words are words that help connect other words but don’t contribute to the meaning of a sentence, such as the articles ("a/an," and "the"), conjunctions like "and" and "but," prepositions like "in" and "of," and pronouns "he" "she." Increasing the density of these words in ChatGPT content can help humanize it, because AI tools lean heavily on content words and minimize the use of functional words. This makes their content more concise to read but it also makes it more predictable and less human-like, since human beings naturally use more functional words.
On the other hand, a higher density of functional words reduces AI detection and makes your content more natural to read. That said, work to add more functional words to the content but don’t force too many of these or the content will become clunky and awkward to read.
4. Adjust the Tone
Another prominent AI pattern in AI-generated text is a generic, formal, and neutral tone. ChatGPT can write in a different tone if asked, but it mostly prioritizes a neutral and formal tone to sound objective and respectful in content, which helps prevent bias and unwanted funny language. However, this neutral tone comes with a lot of robotic terms and expressions that make the content more predictable for AI detectors. That’s why you need to adjust the content’s tone. Make sure it is not overly formal and neutral like that of ChatGPT’s. Ideally, opt for a casual or semi-casual yet respectable tone.
5. Vary Sentence Length
Sentence length is yet another common AI pattern that makes your writing feel robotic. In AI writing, most sentences exhibit similar length. Almost all sentences have a medium to small length, partly because they follow a similar structure. The uniformity in sentence length across the text is because of low burstiness in AI-generated text, which is a metric that measures the variation in sentence length and structure. If a text has low burstiness, it will appear as AI-written to AI detectors and may be flagged as such. You can prevent this flag by simply varying sentence length of the text. Here’s how:
Splitting Sentences: Split some of the sentences to form shorter sentences. Likewise, you can separate two clauses in a single sentence and turn them into two smaller sentences separated by a conjunction.
Combining Sentences: Combine two sentences to form a longer and more complex sentence.
Alternate between short, long, and mid-length sentences for maximum variation.
These five techniques are effective against all AI detectors. Using them properly is the key to humanizing ChatGPT content without compromising its quality.
Conclusion
ChatGPT content is easy, but it needs to be humanized if you want to avoid AI detection. There are a few ways to humanize ChatGPT text but each method has its own pros and cons:
Humanizing Using a Text Humanizer: This is the quickest method to humanize AI text. It is effective, but it doesn’t give you much control over the output text.
Humanizing via ChatGPT: This method is fairly easy but can be tricky. To do this, you need to engineer a lengthy prompt, asking ChatGPT to humanize the text. However, it is the least effective method because chatbots aren’t good at humanizing content.
Humanizing Manually: Manual humanizing is the most effective method of the three. It also gives you more control over the text and its quality. However, it is the most challenging of all. You have to understand common AI patterns and use humanizing techniques to make the content read as natural and human-written as possible. Manual humanization can also be quite time-consuming.
Whichever method you use, remember to check the content for AI detection afterwards to make sure it is adequately low.
