ChatGPT is almost two years old now, relatively young compared to other transformative technologies that have shaped our lives. However, its impact has been nothing short of revolutionary. It has influenced every aspect of our daily lives, from education to shopping. For those of us in education, the question is no longer whether to use ChatGPT, but how to leverage it in a way that is ethically and academically sound.
It is true that while some students effectively use their potential, others do not. This is particularly evident in written assignments. With a single prompt, ChatGPT (or any other ai chatbot) can generate an entire essay, often indistinguishable from human writing, at least to the untrained eye.
However, anyone who has interacted with these chatbots for a long time will tell you that they have developed a great ability to quickly and easily detect ai-generated content. For example, it takes me just a few seconds (certainly less than a minute) to determine whether an article is written by ai.
I've been using ChatGPT to help me with my writing tasks, particularly for blog posts, and I've become intimately familiar with its language and the way it generates content. I've even written an entire book on it. amazon.com/dp/B0D9SWSQWM” target=”_blank” rel=”noreferrer noopener”>ChatGPTdelving into its nuances and understanding how it constructs responses. This deep familiarity has made it almost second nature for me to spot ai-generated text.
Let me clarify one important point: I am not against using ChatGPT as a writing tool, as an editor or research assistant. What I am against is using this technology to replace you and the creative work you do. What I always recommend, especially for writing projects, is to do your homework first (write your initial draft) and only then use ChatGPT to refine it, correct typos, improve grammatical structures, and improve the overall flow of your style.
The key is to develop your ideas and let ai help you refine them, not the other way around. ChatGPT is not a thinking human; when you ask it for something, all it does is pull information from its database and online content, and piece the information together to produce an answer. The risks of plagiarism, inaccuracies, and hallucinations are always present, and I've talked about this in detail in amazon.com/dp/B0D9SWSQWM” target=”_blank” rel=”noreferrer noopener”>my book.
One of the most effective ways to use ChatGPT is as a study assistant, but unfortunately not all of our students use it this way; many will rely on it to get all their work done. The question all educators are now facing is how to detect ai-generated content in student assignments and work.
As you may already know, there are ai checkers that claim to identify whether a text is ai-generated, but the accuracy of these tools is questionable at best. They often miss the mark, especially when evaluating content written by people whose native language is not English, but they are still useful as one of the many tools in our toolkit. While not foolproof, these checkers can provide a starting point for further investigation and help flag potential cases that require further analysis. However, they should be used in conjunction with your own experience and intuition, rather than relying on them as the sole method of detection.
Speaking of experience and intuition brings me to the most effective method: YOU. As educators, we know our students, their academic abilities, and their way of thinking. This insider knowledge is invaluable, and when combined with our understanding of ai robots, puts us in a strong position to detect student use of ai.
Whenever you interact with conversational ai models, pay close attention to the language they use. The more you use these tools, the clearer the patterns of language structures will become. You will notice the recurrence of certain vocabulary and grammatical structures, and over time, you will develop a linguistic intuition that will allow you to quickly detect whether a particular statement or paragraph is ai-generated. This linguistic intuition is your most powerful tool for detecting ai use by students, and is far more valuable than any ai checker on the market.
As a language researcher with a PhD in critical discourse analysis, now fully immersed in educational technology research (specifically, artificial intelligence), I can’t help but pay close attention to the language generated by ai bots. I approach it with a critical eye, always looking for patterns and perspectives to better understand how it is structured.
And yes, ai language follows patterns; recurring linguistic structures are evident, and through my intensive interactions with ai chatbots, I have been able to capture and document them.
Common features of ai-generated text
In addition to the formal expressions you’ll explore below, I’ve identified several general features that often characterize ai-generated text. These can serve as useful guidelines when addressing the problem of students submitting ai-generated content as their own. These features include:
1. Repetition and redundancy
ai-generated language often adheres to a predetermined structure, resulting in the repetition of similar linguistic and syntactical patterns throughout entire paragraphs. As a result, ai-generated text can sometimes feel monotonous and lack emotional depth or variation.
2. Verbosity
ai often offers detailed explanations to make text sound more understandable. For example, instead of saying “The sky is blue,” an ai might output: “It is widely observed that the sky has a blue hue due to scattering of sunlight by the atmosphere.”
3. Generic statements
A recurring problem with ai-generated content is the use of generic and sometimes vague statements. These phrases often lack the specificity and context needed to be truly insightful. For example, statements like “technology is changing the world” or “education is important for everyone” sound meaningful, but they don’t provide depth or detail.
4. Use of the passive voice
Again, this aligns with what we said earlier about the lack of emotional depth in ai language. The use of passive voice often hides agency and obscures the author of the action, making the text seem impersonal and lacking a personal voice and unique perspective.
5. Formulaic structures
ai often follows predictable patterns when writing, such as starting paragraphs with a general statement, continuing with supplementary details, and ending with a summary. This can make text seem repetitive.
6. Excessive use of clichés
ai-generated language often contains overused clichés and common phrases like “a step in the right direction,” “the tip of the iceberg,” or “only time will tell.”
7. Syntactic and spelling perfection
ai-generated texts are typically free of typos, punctuation errors, spelling mistakes, and grammatical inconsistencies. While this alone is not a definitive indicator that a text is ai-generated, it can be a red flag in an educational setting. When a student submits impeccable work, it can lead to further scrutiny.
Commonly used words in ai-generated text
Below are some commonly used words that I was able to identify from my analysis of the piles of ai-generated text I worked on over several months. As an educator, you can use these expressions as indicators to determine whether a student’s assignment may be ai-generated. However, remember to use your intuition and common sense: these indicators are meant to help, but they do not necessarily confirm ai generation. While they suggest a higher likelihood, it is important to take them with a grain of salt.
Final thoughts
I want to highlight that the insights I’ve shared above stem from my own informal research, conducted through extensive interaction with ai chatbots over the past year. Drawing on my experience as a language researcher and my background in educational technology, I have sought to unravel the patterns of ai-generated language. My goal is to equip educators with the knowledge they need to independently identify ai-generated text. These insights could also serve as the basis for a more rigorous research paper for those interested in exploring this topic further.
