Why You Might Want To Paraphrase Most Of What You've Written With ChatGPT
It’s no secret that ChatGPT has been actively disrupting the writing community. However, for such a robust system, I’ve never been satisfied with a ChatGPT response and I rewrite most of them. But, for what reasons exactly?
John Angelo Yap
Updated October 18, 2023
Writing with an AI looming over your shoulder.
Reading Time: 11 minutes
As someone who was born on the tail end of the 90s, all the knowledge I had about artificial intelligence came from Terminator’s Skynet — which is probably why I was under the impression that every machine is out to exterminate us.
That is, until JARVIS from Iron Man came along. Beyond the light beams and the giant mecha robots, I distinctly remember telling myself how useful it’d be to have a machine obeying your every command and answering all your questions.
Hard to believe that I’m staring into that technology now every time I open ChatGPT. It’s rare to see a piece of technology go from an idea to actualization in such a short period of time.
The major difference between JARVIS and ChatGPT is its medium. Whereas the former accepts voice commands from Tony Stark, the latter communicates through text. While it’d be nice to have a voice-activated AI, written interactions in ChatGPT do present us some unique opportunities, particularly in creating content. That said, after using ChatGPT daily, I maintain that:
ChatGPT is great, but it is also greatly flawed. And that’s most evident in its writing, which I’ve always had to paraphrase.
What do I mean by that? Well, better stay tuned.
AI in Education
As much as we can marvel about ChatGPT’s technological prowess, there’s an understated quality of it that’s just as responsible for the AI revolution: accessibility. Imagine, ten years ago — would you have believed that there would be an AI in your PCs, phones, and even in your smart home devices?
And the place where it’s most prominent? Classrooms.
Just like search engines before them, AI tools were always bound to change how we approach education. Truth be told, it already has. In fact, more than 89% of students confess to using ChatGPT for their schoolwork. There are a few that use ChatGPT as a quick way of condensing information from the internet. However, the majority of students using ChatGPT have one goal in mind: get ChatGPT to write their essays.
ChatGPT has left teachers baffled as they try to navigate this new age of AI. How can you ensure student integrity when one single prompt can complete your assigned tasks? So, they turn to AI detection tools such as TurnItIn, looking to end this dilemma. The thing is, even these detectors aren’t perfect due to many reasons.
The biggest of which is this: Did you know that, whether or not you use ChatGPT, you can still be flagged by AI detectors? This brings us to...
False Positives and False Negatives
Have you ever made an assignment that was tagged as AI-generated even when those are your words on the paper? Several students have.
This was the case for Louise Stivers and William Quarterman of UC Davis. Both exceptional students whose academic performance plummeted due to being unjustly flagged by AI detectors. Fortunately, they were able to recover their grades by arguing their case. But, not all have that luxury.
This is what you’d call a false positive: human-written content that’s flagged as an AI. On the other hand, when an AI creates a piece of text and it somehow passes detectors, that’s a false negative. And these aren’t just isolated cases or anecdotes — it happens more than you think. For that matter, it took me only one attempt to replicate these errors.
So, let’s set the table. Let’s use Content At Scale as our AI detection tool. For the false positive test, I’ll be using one of the paragraphs from my thesis; and, for the false negative test, I’ll generate content from ChatGPT. Here are the results for the false positive test:
Now, for the false negative test:
With these results in mind, do you still trust the accuracy of these AI detection tools? Personally, to an extend although I generally don’t. But, it doesn’t matter. Until there’s a better solution, educational institutions will use AI detectors to catch copy-pasters.
This is why you should always paraphrase content from ChatGPT if you’re going to use them on your schoolwork. After all, AI detectors can make a mistake and flag human-written content — what makes you so sure you’re exempt from that?
What is AI Hallucination?
I’ve spent a lot of time with AI tools, ChatGPT most of all. And do you know what’s my biggest realization about them?
They’re not as intelligent as we expect.
Sometimes, I’d ask a question about a specific topic and the answers it gives me doesn’t fit my prompt. Sometimes, it would avoid the question.
And, worst of all — it would sometimes give me a completely made-up answer. For instance, let me ask ChatGPT about the old queen of the Philippines called Suman. Here’s the response I got:
What an in-depth response, right? But, here’s the thing: the Philippines was never a monarchy so it never had a queen. Luzon was also never a kingdom itself but it is an island that housed many kingdoms. Oh, and Suman isn’t a name by the way — it’s a popular rice delicacy.
The tendency of AI models to fabricate information is called hallucination, and there’s a chance you might’ve fallen victim to it already.
Why Does AI Hallucinate?
ChatGPT is the very definition of a “people-pleaser.” For better or worse, AI models learn on their own using publicly available information from a dataset. However, they lack the ability to distinguish which information is correct or false. The problem is, the internet’s as wide as the universe, and there are certain things that will get traction despite being wrong.
Despite knowing that there’s a probability it’s false, ChatGPT can sometimes respond with misinformation just to meet your expectations. Hence, hallucinations.
Here are some specific ways to trigger AI hallucinations in ChatGPT:
- Informal Register: Idioms, region-specific language, or new slang after the latest ChatGPT update.
- Leading Questions and Triggered Confusion: Prompts that have a specific answer in mind.
- Long Conversations: Extensive conversations can have a huge influence on ChatGPT and even cause the model to mimic you.
- New Information: ChatGPT’s knowledge base only extends to September 2021.
- Vagueness: Unclear prompts with subjects that can mean multiple things.
What Can We Do About It?
Verify everything. Got an essay about the Spanish Inquisition? Fact-check it. GPT gave you a complete list of newsletters to follow? See if they exist.
You can’t take everything ChatGPT gives you at face value. Paraphrasing your text to reflect accurate information is an essential process to getting the most out of ChatGPT.
Does Google Penalize AI Content?
There’s been an ongoing debate in the blogging community about whether or not Google penalizes AI content. The answer? Not strictly.
Google is steadfast on their demand for quality and engaging content. Good articles are rewarded with better SEO rankings, as they’ve always been, and that includes AI-generated articles.
Per Google Search’s guidelines, you can publish and be rewarded for AI content as long as they’re high-quality and people-first. What that means is that you can’t just take everything from ChatGPT, paste it into WordPress, and expect good returns. Instead, you have to take the questions like the following into account first:
- Does the content provide a substantial, complete, or comprehensive description of the topic?
- Does the content provide insightful analysis or interesting information that is beyond the obvious?
- Does the content have any spelling or stylistic issues?
- If someone researched the site producing the content, would they come away with an impression that it is well-trusted or widely-recognized as an authority on its topic?
The one thing that Google Search frowns upon most of all is content made to attract visitors, instead of informing them.
So, yes — you can use ChatGPT to create content that ranks. However, you must always exercise caution and strictly follow Google’s guidelines. Always verify your sources and paraphrase text from ChatGPT in such a way that appeals and communicates knowledge to readers.
Why ChatGPT Can't Replace People as Writers?
In May 2023, the Writers Guild of America took to the streets to fight for a more livable wage. While this strike is significant, it’s hardly unprecedented — in fact, in 2007, the WGA also successfully protested for increased residual rates.
The one key difference between these strikes? Executives have an ace up their sleeve this time: artificial intelligence.
As a writer myself, I’ll admit that the prospect of being replaced by sentient chatbots scares me a little. But I have one respite from this fear:
ChatGPT may be smarter than me, but it lacks creativity.
I’ve been using ChatGPT and Bard since their release, and it’s pretty easy to notice how poorly-written some of their responses are. For example:
When I asked ChatGPT for a creative essay, what I’m provided sounds like a simple essay ran through a thesaurus. It doesn’t exercise creativity in the same way we do — through playful commentary, length variation, idioms and similes — instead, ChatGPT’s idea of creativity is complexity, which makes sense because it’s a machine.
I don’t frown upon people who use ChatGPT for articles or scripts, but the difference between human writing and AI models is farther than the gap between stars. Only use ChatGPT as a framework and a research tool for your content.
Don’t let it define who you are creatively.
How to Paraphrase ChatGPT Writing
There are no foolproof methods of bypassing AI detectors — but you can lessen your chances of being tagged. Here’s a list of things to keep in mind when you’re transforming AI-generated content:
- Add Idioms and Similes: Adds creativity to an otherwise monotonous content.
- Avoid Common GPT Phrases: Some examples include “Firstly,” “It’s important to remember,” or “Picture This.”
- Create Sentence Variation: Adjust the length of the sentences to create a better flow.
- Don’t Be Neutral: Emotions have a significant role in human writing.
- Trim the Fat: Remove unnecessary information from ChatGPT.
The Best AI Bypass Tools
Don’t have the time to paraphrase? You can always use AI bypass tools to rewrite your content.
You must, however, always consider that these software are AI-powered themselves. They can only mimic and tweak the content, not add anything substantial to it. Moreover, they’re not perfect: it can sometimes give you an output that you still have to tweak.
That said, here are my two recommendations for such tools:
Undetectable AI
With the goal of humanizing AI content, Undetectable AI is one of the best AI rewriters in the market today. It’s capable of changing your text into something that won’t get flagged by AI detection tools whilst retaining its readability.
Furthermore, you can customize Undetectable AI’s output depending on your:
- Use Case: General Writing, Essay, Article, Marketing Material, Story, Cover Letter, Report, Business and Legal Material.
- Reading Level: High School, University, Doctorate, Journalist, Marketing
Undetectable AI starts at $9.99 per month for 10,000 words, and increases to $209 for 380,000 words.
HideMyAI
Similar to Undetectable AI, HideMyAI aims to reduce the likelihood of a text being tagged as AI.
HideMyAI offers more customization options than Undetectable AI, including language, perspective, and generator. However, that doesn’t mean that HideMyAI produces a better output — for a more in-depth comparison, you can read more about these two platforms in our article here.
The cost of HideMyAI starts at $8.25 per month for 10,000 words, and increases to $145.20 per month for 200,000 words.
Some Last Words
You can’t put the rabbit back in the hat. AI is here, and it’s here to stay — and I’d argue that that’s a good thing.
ChatGPT, for me, has been an immensely useful AI writing assistant for research and proofreading. It’s an unprecedented but welcome software that helps streamline the demands of my day-to-day life.
However, overreliance on AI does exist. Not only does this lead to possible misinformation and academic or workplace sanctions, it can also severely handicap your soft skills because it affects your productivity and efficiency.
At the end of the day, ChatGPT is a tool, not an answer. So, use it as such.
Want to Learn Even More?
If you enjoyed this article, subscribe to our free newsletter where we share tips & tricks on how to use tech & AI to grow and optimize your business, career, and life.