With the world entering a new phase where AI is prevalent in many aspects of our daily lives, a place where the issue is at the forefront is inside our schools. An increase in students being caught cheating using AI on assignments has led to teachers reaching for AI detectors to catch this. Although this seems like something that can ensure academic honesty, these detectors are not always accurate and can lead to either innocent kids being caught for cheating or kids who did indeed use AI being able to slide through without being caught.
To put this to the test, I put three historical documents into three different AI detectors — Zero GPT, Copy Leaks and Writers.com — to see if they could truly detect what was AI and what was a real document.
The Declaration of Independence:
The Declaration of Independence was written on July 4, 1776, which famously announced America’s intention to break free from Great Britain. This document, which was so famous and written long ago, is a good candidate for an AI detector. After being put through the three detectors, both Copy Leaks and Writers.com said that the document was human-written. However, Writers.com said it was only 75% human-made and 25% AI-generated. Most notably, Zero GPT determined that the document was 97.75% AI-generated. It’s also important to note that Zero GPT is the first link to show up when searching for AI Detectors on Google.
President Biden’s Inauguration Speech
President Joe Biden gave his inauguration speech on Jan. 20, 2021 in front of a large crowd in Washington, D.C. This speech was detected much more accurately by AI checkers than with the Declaration of Independence. Zero-GPT said that it was 75.67% AI, while Copy Leaks and Writers.com both agreed that it was fully written by a human.
The Gettysburg Address
The last real document tested was the Gettysburg Address into the AI detectors. The Gettysburg Address was given by President Abraham Lincoln on Nov. 19, 1863. This famous speech by Lincoln was given during the Civil War and is known for Lincoln promising a “new birth of freedom.” After putting this document into AI checkers, not surprisingly after the first two documents, Zero-GPT said that the address was 100% AI. Meanwhile, both Copy Leaks and Writers.com once again said that it was human writing while Writers.com did say that it was only 83% human.
Moving on, the next thing put to the test was AI-generated texts using prompts that were made by Chat GPT.
Are Hot Dogs Sandwiches?
For the first prompt, Chat GPT was asked to “write me a paragraph about why hot dogs are sandwiches.” What resulted was this: “Hot dogs can be considered sandwiches because, at their core, they fit the definition of a sandwich: a food item consisting of a filling placed between pieces of bread. A hot dog is typically served in a split bun, similar to how subs or hoagies are served, which are universally accepted as sandwiches. The bread surrounds the filling—in this case, a sausage or hot dog—much like how other sandwich varieties enclose their contents. While the hot dog bun might be connected on one side, many sandwiches, like a lobster roll or a sub, often maintain a similar structure. Thus, classifying a hot dog as a sandwich is consistent with the broader understanding of what a sandwich can be.”
After putting this through AI detectors, none of them detected that this was fully AI. Zero GPT detected only 39.68% AI in the text, while Copy Leaks determined this was human text and Writers.com said that it is 93% human.
AP US History Short Answer Question
To give a more accurate example of what a teacher may see and question on a daily basis, I put the rubric and question of an AP US History Exam SAQ. After analyzing the response, Zero GPT and Copy Leaks both detected 100% AI, but Writer.com found that 70% of it was human-generated.
In Summary
With these AI detectors barely being able to accurately detect whether a text is human or AI-generated, it may be important for teachers to find new methods to detect cheating that would hold students accountable when they do cheat while not accusing innocent students.
