For many of us, writers or otherwise, ChatGPT has fundamentally changed how we think about content—for better and worse. On the one hand, artificial intelligence (AI) has allowed us to easily fill gaps in our workflow, from the pre-writing phases all the way to editing and proofreading.
On the other hand, ChatGPT, Claude, or any other AI chatbots can also do the writing itself. Although they do it quickly and efficiently, it comes off in a way that feels robotic, unnatural, and, in a word, artificial.
While any potential worries that writers might eventually be replaced entirely by AI are overblown, it doesn’t mean that there aren’t some concerns about AI’s influence on the world at large, which is one of the reasons the recent writers’ strike happened. Strike or not, ChatGPT and other forms of AI will continue to be a large part of our lives.
The trick, then, is to find ways to work alongside and with AI. If you’re a writer, this might start with ensuring that you don’t fall victim to any AI detection tools, like AI Detector.
Why Bother Avoiding AI Detection as a Writer?
Sure, there’s some sense to be had with a proclamation such as, “If you can’t beat ‘em, join ‘em!” However, with AI content becoming increasingly prevalent over the last two years, it’s good to ensure that detectors don’t flag your content.
For starters, if you’re a student, the reason is obvious enough: you don’t want to get accused of academic dishonesty, especially if you’re not actually using AI since false positives can still happen. For professional writers, in the worst-case scenario, AI detection might impact your reputation or ability to be paid.
Image credit: GPTZero
On top of that, there’s another factor to consider: it’s exceedingly important that whatever you write stands above what ChatGPT can produce. The best place to start is by ensuring that your content is wholly differentiated.
And if you’re using AI to assist with content—no judgment—these tools can be valuable, but it’s great to ensure that what ChatGPT creates doesn’t read as though it was written by AI.
The best place to start with avoiding detection as a writer is by developing a clear understanding of AI’s writing style so that you can either ensure that your writing differentiates itself or you can effectively edit AI-written content. So, let’s start there.
The Qualities of AI Writing
ChatGPT, despite its natural language and ease of conversation, is still just a program. The intelligence it exhibits is, not to state the obvious, artificial. Computers don’t think like we do; they think in ones and zeroes. And if you take a step back and truly digest what an AI chatbot like ChatGPT produces, you can definitely get this sense.
AI can give us the responses we want in a way that’s easily digestible and full of good information—although sometimes that information can be wrong or false—but it gives us this information in monotonous, robotic output that works well in a utilitarian sort of way—but creative and well-written it is not.
So, let’s explore the distinctive qualities of AI writing, starting with the big picture and working our way down to the more granular.
Personality and experiences
Some writers out there also lack personality and have difficulty inserting a distinctive voice, but AI does, too, and to a much higher degree. With a basic prompt, you’ll receive a basic response that could be compared to the plain bread of writing. It won’t have much in the way of voice, humor, or any sort of clever wordplay—just information on a page.
Humans tend to draw from their own experiences in their writing. Because AI doesn’t have actual experiences, on many levels, it’s just regurgitating what it’s been trained on in an easy and accessible fashion.
On the subject of prompts, often, you can substantially add some personality to a chatbot’s response with some prompt tricks (which helps the chatbot’s application of other good writing, too). However, while this will certainly provide a much better response, one that will read less like an AI wrote it, nothing you can put into the prompt—barring any really lucky results—will completely avoid AI detection. You’ll probably just bring your text from 100% AI to 50% or 75% AI.
Reduced variance
Often, you’ll hear this called “burstiness.” It refers to the overall variance of sentences and paragraphs, specifically length and structure. AI has a highly monotonous style, with many of the sentences being constructed similarly and the sentences and paragraphs being, on average, the same length.
You can imagine the result of reading a few thousand words in which all paragraphs and sentences are roughly the same length. It gets painful—and fast.
Limited randomness
Similar to variance, randomness—also called “perplexity”—refers primarily to the word and phrase choices. In an AI-written piece of content, the variance is lacking. The reason for this, though it’s a bit reductive, is that chatbots are, in a manner of speaking, highly advanced autocomplete engines.
Image credit: Scribbr
No matter what it’s writing, AI chooses the most logical word and phrase that would naturally come next. So, the lack of burstiness can make AI writing monotonous and repetitive, and without perplexity, AI writing is dull. In short, AI writing rarely presents anything exciting or unexpected.
How Do AI Detectors Work?
While, by this point, you can most likely guess the best way to avoid AI detection—by simply not writing like AI—it’ll also be helpful to explain a little bit about how AI detectors work, which will reinforce your skills at avoiding writing that could be flagged as having been generated by AI.
Believe it or not, AI detectors are built in much the same way that ChatGPT, Claude, or any other model, are built—with the power of a large language model (LLM). LLMs are basically advanced AI models that are trained on vast amounts of human language. We’re talking tens of billions of words. Along the way, developers help to fine-tune these models to get the desired results.
Image credit: OpenAI
Only, in the case of detectors, the models are trained on the qualities we’ve discussed about AI writing, specifically burstiness and perplexity. To this end, they are trained on both vast amounts of AI writing as well as vast amounts of human writing so that detectors are able to discern the qualities of each.
During this process, much like for chatbot models, a developer will fine-tune the results so that not just any dull and monotonous writing gets flagged by detectors, but specifically that which is written by AI.
This training by the LLM is facilitated by a complex neural network, which makes up the framework on which the LLM is built. Not to get into too much of the nitty-gritty, but these networks are how the models are able to be trained. They’re able to determine when one piece of text has features of being AI written compared to another. Make this comparison a few million times, and the model is pretty good at comparing AI writing to human writing.
So, How Do We Avoid AI Detection as Writers?
Now that we’ve laid the groundwork on what AI writing looks like and the inner workings of the detectors, let’s talk about avoiding detection.
For starters, if you’re a talented writer, you probably don’t possess many of the qualities inherent to AI writing. You most likely have plenty of voice, draw from your experiences when appropriate, and maybe even add a dash of humor to your writing. In addition to that, chances are your sentences and paragraphs have just the right amount of perplexity and burstiness.
Next to that, if you’re either trying to pass off AI-written content, or have mixed it into your own writing, then really, the only way to avoid AI detection is with a fair amount of revision and editing.
If your number one goal is to bypass tools like AI Detector, you can focus solely on improving your writing’s perplexity and burstiness, which is what these detectors primarily focus on.
To increase the perplexity of your sentences, one place to start is by varying your word choices—use synonyms, mix in more complex or unexpected terms, and don’t be afraid to get creative with your phrasing. Thesauruses are always handy, but take heed not to overdo it. Though that might avoid AI detection, it’ll detract from the overall readability and flow of your writing.
Where improving sentence structure is concerned, avoiding detection revolves around breaking AI’s robotic pattern. Try flipping clauses, starting with questions, or using more nuanced transitions. However, even here, be careful not to overdo it. You want to engage readers, not confuse them with meandering sentences that resemble an M.C. Escher drawing.
When it comes to burstiness, focus on the rhythm and flow of your text. AI has a habit of keeping sentence lengths uniform, which feels unnatural. So, extend a few sentences, maybe combine a few others. You can also shorten some that might have unnecessary information or split them into two shorter sentences.
In fact, short sentences actually have tremendous power. They can break up the flow in intelligent and impactful ways. So, they’re great not only for increasing burstiness, but also for adding a fair amount of human style to your content. Ultimately, your end goal is to ensure that your writing resembles an actual conversation rather than a chatbot regurgitating information.
Although avoiding detection is one thing—your main goal should be to connect with readers as best as possible.
Elevating Your Writing Beyond AI Text
Doing some light rephrasing and sentence rearrangements will allow you to partially or fully avoid AI detection—at least allowing you to go from 100% AI to 30–50% AI.
However, with as prevalent as AI-generated text has become across the internet, it’s more important now than ever to write content that truly captures readers in a way that chatbots can’t, especially since simply avoiding detection will only serve to edge you above the AI competition.
So, if you’re looking to grow a platform, or at the very least, create content that readers actually enjoy from start to finish, it’s important to write better than AI. Fortunately, we can use the information from earlier as a blueprint for how to achieve this lofty goal.
Voice and style play a large part in this. Imagine that you’re having a conversation with the reader, so write how you’d speak to a friend—while also using better grammar, of course.
It’s also worthwhile to insert your own experiences or anecdotes when you can. Since AI only really regurgitates info, if readers can feel like you actually did something or are an expert on the subject, then they’ll be more likely to listen to what you’re saying.
Expressing your own feelings and opinions is a great way to elevate your writing over that of AI, too. Chatbots simulate feelings, but they don’t feel anything the way we do, nor do they really have any opinions on any subject. Although not every topic has room for these qualities, they can be instrumental when appropriate.
Nail just one of these ideas, and you’ll have writing that won’t just avoid detection, but will meet a high-quality bar.
Using AI Tools To Help Avoid Detection and Enhance Writing
As you can easily imagine, AI is not necessarily something to be feared or avoided. The truth of the matter is that it will most likely remain an integral part of our lives for the foreseeable future. So, the best way to coexist with AI is to learn to use these tools to your advantage.
For starters, the same AI detectors we’re trying to avoid can be a huge asset in achieving that coexistence. That’s because when you scan something with a detector, it tends to highlight any problematic areas. So, if all you’re trying to do is narrowly skim by, then you can just focus on which sentences or paragraphs are flagged by the AI detector.
Beyond that, as we mentioned at the start of this article, ChatGPT can be integral to the writing process—and that alone bears repeating and expanding upon. AI can help us with research whatever we’re writing about—although be on the lookout for inaccurate info and take the initiative to double-check facts yourself.
Outlining is also made easy by AI. If you’re writing about a topic and want to make sure you’re covering all your bases, you can have a chatbot help you fill in any gaps. You don’t necessarily have to write about everything that ChatGPT suggests, but it might help you catch anything you might’ve missed or not considered.
Then, once you’re done writing, these chatbots can also help you with proofreading and revision. Although it’s better to have AI instruct you on what to do rather than letting the tool make any edits itself since that runs the risk of causing writing that could be flagged as AI-generated—and that’s counterproductive to your goals.
Ultimately, no matter where you are along the writing process, AI can give you actionable feedback that will help enhance your content. Even tools like Grammarly, which have also been built with AI, can be tremendously helpful during the writing process, although be careful with its enhancement features since the writing that it adds can be flagged, as well.
The Bottom Line
As silly as it might sound, the best way to avoid AI detection is by not writing like a robot. And with any luck, this article has provided a firm foundation to avoid doing just that. After all, as powerful as AI is, humans have—and will probably always have—the upper hand when it comes to writing.
However, if you’re unsure about how your writing fares compared to AI, tools like AI Detector can be instrumental in letting you know just how similar your writing is to that of a chatbot.
Though AI can prove to be a tremendous benefit to anyone writing content about any given topic, having your text flagged as AI can affect anyone, most notably students trying to avoid accusations of academic dishonesty. Also, professional writers can use these tools to ensure they maintain a spotless reputation in their careers.
