ChatGPT and AI Detection

It seems that lately, talk of artificial intelligence (AI) generated content has been everywhere. Much of the buzz is due to the launch and sudden explosive growth of OpenAI’s ChatGPT, which statistics show gained a million users in its first week. 

The proliferation of AI tools, like ChatGPT, has caused consternation in a number of areas and industries in which original human-generated content is valued. These include (among others):

• Education
• Content marketing
• Publishing
• Journalism
• Law

The central questions being hotly discussed are: “How can we differentiate between AI and human-written text?” and “How might we detect AI-generated content?” 

Let’s explore.

How does AI-generated text work?

To create machine-generated content, programmers “feed” the artificial brain large pools of data in a process called machine learning. Becoming Human: Artificial Intelligence Magazine explains: 

“…first it is fed a large dataset of text, such as articles or conversations. It then uses this data to learn the patterns and structure of language. Once it has learned enough about language, it can generate its own text based on a given prompt or topic.”

The idea is that, over time, the AI will learn to replicate human thinking and natural language processing well enough that the reader will be unable to determine that they are reading AI-written content. 

This can be appealing, especially when a person needs to produce content at scale (like blog posts), write about a topic they don’t know much about or complete a paper in a hurry. There are also substantial drawbacks to AI-powered content:

• It is often factually inaccurate and can include outdated data.
• It often features poorly constructed phrases or inarticulate thoughts.
• The technology does not allow for nuanced thinking in many cases.

It’s helpful to remember that AI is only as good a writer as technology and machine learning allows it to be. While ChatGPT (and others, such as Bloom, Jaspr, and Rytr) are effective tools with great potential, none can yet replace a human writer.
The drawbacks and questionable ethics (as discussed below) make AI detection an essential tool wherever original writing is valued.

Why does AI Detection matter?

For some, the ethical questions surrounding AI writing may seem insignificant. After all, is content generated by ChatGPT plagiarism if it came from a single (computer) writer? But in education and enterprise, there are substantial and vital concerns that include both AI plagiarism and the utility of writing itself. 

AI and Education

In education, it is the provenance of the institution to teach students to write and to complete original research. It is not uncommon for writing to be included across the curriculum because it is a highly effective way for students to demonstrate their understanding of a topic.

If a student employs an AI tool to generate their writing, they may be able to bypass the curriculum altogether and fail to demonstrate either their research or their writing skills. Unless an educator can verify a strong likelihood that a human writes a student’s paper, they cannot effectively demonstrate that the student has met the institutional learning outcomes.

Because it’s impossible to attribute AI writing to the student submitting the paper, it is evident that turning in work created by AI is plagiarism. As we’ve discussed on our website, “When someone uses the work of another writer or artist without properly citing the source or giving credit, that’s plagiarism.” Hence, barring direct attribution to the AI itself, the definition fits. In our blog, we’ve spoken extensively about plagiarism’s ethics and the problems it creates in the academic sphere.

AI and Enterprise

In the business world, content, and copywriting are the most likely places to find AI writing. It may appear to be a viable tool for marketing copy. Still, as with using AI in education writing, there are also substantial drawbacks to using AI content for content marketing writing.

Suppose, for example, you hire a content writer to create blogs for you, and they deliver blogs written by AI. Have you indeed received the service you paid for? If you are a business owner using AI yourself, the ethical questions may be diminished, but they are not altogether nonexistent. Can you be sure that the AI didn’t directly plagiarize from the data pool?

Perhaps more pressing is that, by all indications, using AI content diminishes one of the main advantages of good content in the first place: search engine optimization-related ranking potential. Suppose you are creating copy to increase your rankings on Google. In that case, it’s vital to know that, as Search Engine Journal has noted, “Google’s Search Advocate John Mueller says content automatically generated with AI writing tools is considered spam, according to the search engine’s webmaster guidelines.” That means, if you’re trying to drive traffic to your website, using AI-generated content could hurt more than it helps, at least for now.

If we are to believe Google, and unless things change dramatically, AI content will not increase your rankings based on Google’s most recent algorithm updates. Does that mean you should avoid AI writing tools like ChatGPT altogether? Perhaps not. However, it does mean that if you choose to use AI, it is in your best interest to deploy a powerful AI detection tool like Copyleaks.

How does an AI detector work?

There are a few ways that AI detection tools operate. The first is looking for indicators within the text, such as

• Linguistic patterns
• Repeated words or ideas
• Structural features

But some AI detection tools, like Copyleaks, are equipped to look a little more deeply. We’ve deployed an AI tool ourselves to help combat AI plagiarism. As Alon Yamin, CEO and co-founder of Copyleaks, explains, “With 99 percent accuracy, we’re able to combat the dark side of AI by uncovering AI-digital DNA crumbs that are left behind, which only sophisticated AI like what we’re using is capable of detecting.”

As we have noted in our AI detection launch press release, “Copyleaks has expanded its content authentication platform with the most comprehensive solution available through customizable API and LMS integrations, in addition to browser plug-ins for individuals, to verify if the content is original or generated by an AI chatbot.”

To learn more about our AI detection capabilities, book a demo today!