Tools like ChatGPT and QuillBot are becoming more and more common in the field of content creation as artificial intelligence takes over the web. Together, these AI-powered tools can rewrite text and generate responses to questions, potentially saving writers and content creators countless hours of time and effort.
But the issue still stands: Can ChatGPT combined with QuillBot really produce indiscernible, AI-generated content? Can these tools reliably and sustainability create articles that are identical to those created by humans? The short answer is somewhat, the longer answer is that unique human writing is several levels above what AI can do – both in terms of creativity & advanced language.
As a content creator myself, it's important to note that I do not condone the use of AI copywriting tools as a substitute for human effort and creativity. While these tools can certainly be useful creativity boosters, they should mainly be used to supplement the efforts of experienced writers — not replace them. Google will also boot your website if you're trying to start a blog with a bunch of AI fluff.
Google uses extremely advanced Natural Language Processing models, which are designed to detect and hinder AI-generated content from rising to the top of their search engine. But there's a difference between trying to scheme a billion-dollar company and utilizing a few shortcuts to speed up the company progress report you haven't started that's due at the end of the week...
That being said, I wanted to do some experimenting. I've been reading a bit on Twitter about how ChatGPT and Quillbot can be used to create human-level content, so I decided to give it a try.
What Does AI-Detection Even Mean?
Machine learning algorithms are used to analyze content and look for characteristics that are typical of AI-generated output. It works somewhat like reverse engineering based on matching words to predicability. Since AI-detection basically detects the chances that user text would also have been produced similar to if you asked a robot to answer the same prompt, the higher chance user text is similar to what AI would have produced based on the previous word, the higher chance something is AI.
See the green text? That means the following word was what the AI would've predicted (as one of the top 10 next words) if it had to write it by itself. Human writing is very versatile & creative, AI really isn't. Creativity is what makes organic writing unique, not being able to predict patterns is the most measurable factor in determining the odds something is natural writing. In a weird oversimplification: predictable human text can easily be produced by AI – unpredictable text can't be. In the following example "I was stationed in ___", Afghanistan is the 3rd most likely word based on what the data the AI was trained on (GPT-2). There are plenty of ways of checking if text was written with AI, but they all involve this predictability method.
Additionally things like language usage patterns, strange phrase constructions, and other abnormalities more likely to be produced by a machine than a human increase flags in readers that don't look natural, giving content more odds to be either poor writing or generated by a bot.
Using ChatGPT & QuillBot to Create Unique Content
Since we're on the topic of generative AI, I decided to stick with that theme for the fake article we'll be constructing. I went ahead and typed this prompt into ChatGPT and got an output that looks very AI-like. The more you work with these tools, the easier it is to detect how "engineered" some of these responses are:
After this, I went ahead and threw it into Originality. To date, Originality seems to be the best AI content detection platform for industry-level content. Anyways, here's what I got:
With great shock (sarcasm), I was astonished to see Originality marking our text as being 100% AI-generated. Ok... so you clearly shouldn't use this for anything professional, you'll look like an NPC in a video game. It's not necessarily bad or incorrect content, but it looks like something a high school junior scrambled together from Wikipedia articles about 2 hours before it was due for class.
The next step is to make it uh... not look like that. I threw the first paragraph into QuillBot to start and after a few seconds of tweaking synonyms, cleaning up ugly syntax, and making sure the whole article was cohesive, we got to this:
Much, much better. And once again I did it with the second paragraph and got this:
After throwing the newly produced content into Originality, we've dropped our AI detection rate to a mere 52%. And this is straight out of the box, I haven't edited outside of QuillBot yet. Here's what it looked like after throwing it in Originality and GLTR.
Modifying the ChatGPT + QuillBot Results
Now here's where the human comes in. You should take this content (that was already on the right track to being "undetectable" and go through to proofread it for readability. Regardless of AI detection scores, if a text isn't readable, nobody is going to read it (at least not for long). This standards holds true if you're writing for Google, academic papers, or business reports.
The quality of your writing will always matter more than any numerical score a tool assigns it. I mean, would you continue reading this article if it was bad? After making changes and editing for readability, I ended up getting these scores with Originality:
Now honestly, you could've just written these two paragraphs yourself in the same amount of time, right? Well the two best use cases I would suggest is for a) idea generation and b) longer instances of text. Some of the ideas given in the paragraph I probably wouldn't of just made up while brainstorming. Combine this with paragraphs of content and you could be saving yourself tons of time.
How Good Is It?
Now this obviously isn't a perfect method, but I think it's a step in the right direction if you want to de-AI certain content. I think the best use case of this would be for short-form headline or website copy, not for writing full essays and articles. It'd be a huge pain with longer content, and it doesn't really do an amazing job. There's something about the flow of the paraphrased writing that just doesn't flow like it would if you just wrote that paragraph yourself.
Once again, I'm not advocating for the use of AI-generated content to produce complete papers for you, but I think mixing ChatGPT, QuillBot, and Originality can be some powerful tools in a content creator's arsenal to optimize their workflow. It can produce unique output that could potentially save hours of manual work.
Remember, AI detection basically means giving text a predictability rating! Natural writing cannot be replaced by bots since the human "data set" (your brain) uses experiences, memories, education, etc to wrap every thought you have into words, sentences, articles... you get it!
How Could You Use This Content?
Personally, I think this should only be used to develop short-form content or product outlines if you're going to use it for some production. ChatGPT is the most powerful tool out of these 3, but mixing it with QuillBot and checking with Originality could yield some interesting results. Here are the best ones I came up with:
- Generating eCommerce product descriptions: ChatGPT and QuillBot could be used together to generate product descriptions that effectively highlight the features and benefits of the product in a way that is both persuasive and compelling to potential customers. ChatGPT could be used to generate the initial text based on the product's specifications and target audience, while QuillBot could be used to fine-tune the language and ensure that the descriptions are well-written and free of errors. This process would save HOURS of time and resources for eCommerce businesses, especially once ChatGPT is prompted with the correct keywords and parameters about your product/store. It's a no brainer – especially if you're setting up a store with a ton of products.
- Outlining website blog posts: These tools could be used together to generate outlines for blog posts on a service-based website. ChatGPT would generate the initial outline based on the topic and main points of the blog post (which you would prime with as much information as possible), while QuillBot would be used to refine the structure and ensure that the outline is logical, well-organized, and includes keywords that are actually engaging. This process could be especially useful for businesses and organizations that publish frequent blog content, especially if the copywriter is well-versed in the topics they'll be writing about. Save hours of time, resources, and money by outsourcing the outlining process to these tools. If you reverse-engineer ChatGPT, you could actually write the entire body section of your article, then ask ChatGPT to create an intro and conclusion around it. Paste that result into QuillBot to further customize & turn your content more creative.
- Creating engaging content for a YouTube script: ChatGPT and QuillBot could be used together to create engaging content for a YouTube script. ChatGPT would of course write the script on the topic and tailor it toward a specific target audience, while QuillBot could be used to adjust the tone. This process could be particularly useful for YouTube creators who need to produce high-quality video content on a regular basis, as it could save time and resources by automating the creation of scripts. If you're a script writer who mass-produces short-form content for platforms like YouTube Shorts or even TikTok, this could become an effective rapid publication strategy (just remember to fact check).
As artificial intelligence evolves, tools like ChatGPT and QuillBot are being thrown around more frequently in the realm of content creation. These AI-powered tools can rewrite text and produce questions and complex answers – significantly lowering the time and work for many content creators. The power of these tools to reliably generate articles indistinguishable to those written by people is still debatable. I don't think we're there yet. As impressive as they are, these tools shouldn't be utilized in place of experienced writers, but they can be beneficial in helping them. Google also uses sophisticated Machine Learning models to detect AI-generated content and will flag websites from search results if it violates their terms of service. While these techniques can be blended to produce creative, human-like content, it's crucial to maintain blogging ethics or you may risk your online copy getting docked by search engines when they find out. It will be very interesting to see the development of company copy & blog writing over the next few years!