Ways to detect AI-generated educational work such as ChatGPT

The education system is endangered by AI🏚️

It is no secret, indeed it has even gone viral, that tens of millions of students around the world are cheating within the education system, mostly in high schools and universities. And teachers lack the tools to detect this cheating.

But fortunately solutions are already available. And in this article I am going to give you 3: the one they are using in the United States, the best one for chatGPT, and a third tool that is not used for this purpose and has gone unnoticed but is really useful in the educational field.

Today, there are millions of students who are cheating, handing in AI-generated work and even taking exams with its help, thanks to the fact that many of these assessments are online.

The best tools for detecting AI-generated text 🛠️

  • GPTZero: It is a tool capable of detecting if a text has been generated by AI, it is probably the most used tool in the US, it detects ChatGPT, GPT4 and other artificial intelligences. But it does not always detect them, especially GPT4. It is free and if you register you can use it many times, for extra features, more usage, API, etc. you have to pay. Go to GPTZero.
  • OpenAI detector: It is offered by OpenAI itself, the company behind ChatGPT. It’s free but you have to register. In my experience it usually detects ChatGPT, but not GPT4. Visit OpenAI Detector.
  • Copyleaks: It is a company dedicated to detecting copied works and articles on websites copied from other sites, even if it is in another language. With the spread of AI-generated text, they have launched a tool to detect it. I don’t recommend it, because unlike the rest of this list, it gives a lot of false positives. So many that in my opinion it is almost useless. However, I have been using it for years to detect copied text and it is very good. Copyleaks Website.

These tools, especially the first two, are useful for detecting AI-generated text such as ChatGPT, but they are not perfect and will not always detect it, so many exams and papers generated by artificial intelligence will be missed, and as I said in a previous article, it is a fight that is ultimately lost, it will not be possible in one or more years to detect which text has or has not been generated by artificial intelligence.

Therefore, I propose a different method, instead of detecting whether a text has been generated by AI, let’s check whether the student has actually done his or her work or exam.

Detecting cheating in exams and papers that have been generated by AI 🍣

The tool I propose is ChatPDF.

What is ChatPDF? It is a free online tool where you can upload texts and files in PDF and other formats up to 120 pages, if you want to upload a larger file you have to pay for the application. Once the file has been uploaded, you can chat with the AI about the article.

It is neither practical nor possible to memorise the work of dozens of students to ask them about their own work to see if they know what they have written or if they have generated it by AI and do not even know what they have done.

Thanks to this tool we can ask students questions about their own work and check if they are getting it right.

How? Very easily, imagine that your students have done a paper on history, but one has done it on one part of Japanese history, another on another period of Japanese history, another on Japanese history but with the focus on economics, another on the history of Mexico or Spain, etc. It would make no difference whether the papers were on philosophy, history, law, etc.

You take that work, upload it to ChatPDF and ask him to ask for example 5 relevant questions about the content of the PDF, then ask him for the answers in the PDF to those questions. And with that information you now ask the student the same questions, see what he/she answers… if he/she has created it, he/she should know what he/she says, explains and concludes his/her own work.

ChatPDF Website.