How to Check If Someone Has Written with AI.

From the beginning of time (or at least since the 20th century), teachers have always been nagging students about academic honesty. Before, this was just about copying down exactly what a book or website says, which is quite easily checked and eradicated. But in the past decade, a new cheating method that students have at their disposal has emerged: AI. 

A symbol of technological development, the pinnacle of artificial creation, and the perfect tool for students to cheat. AI can generate text that can fulfill creative writing criteria in seconds, analyse and completely destroy any and all math problems, and contains a plethora of information that it can organise and compile much faster than any human can search the web. Now this obviously is a huge problem, at least for teachers, and so many different methods have been created to try and sniff out any AI generated text, whilst students have been trying ways of hiding the fact that their essay is not theirs. This underground war has been going on for quite some time, and although I am inclined to join the student’s side, Mr Coe is reading the text over my shoulder right now, so I’ll have to take the teacher’s side on this one. 

So, here are my researched, refined and 99.99% effective AI detection methods:

1. Use an AI checker. As simple as that. Literally a tool available in order to detect any use of AI. However, AI checkers take quite some time to analyse the text and generate the result, especially if a teacher has to check the essays of many students. Even if you decide to invest this time, AI checkers are also not that accurate: they identify patterns from AI generated text, which can be inaccurate at times, and other times maybe a student just has a very similar writing style to AI. AI checkers usually only check for texts with less repetition, slightly botched contextual information and paragraphs with similar text length. However, a human could just slightly change word choice, or could show the AI their own writing style and mirror it. This is probably the most accurate method, but definitely has its flaws.

2. Check if the text is too perfect. Obviously, an AI would not make common grammatical mistakes that us basic humans would make. Their grammar and spelling would be completely flawless. This is definitely a good way of detecting AI content at lower grade levels, however when you get to high school/university levels, perfect grammar and spelling are common in order to boost ethos. So this tip is basically useless if you’re a university teacher.

3. Looking at the tone of the text: AI-generated writing tends to sound flat and overly formal, which makes it pretty easy to spot. Then again, didn’t I mention earlier that AI can mimic a student’s tone? The thing is, it often goes overboard—like, if you ask it to include slang, it'll throw in Gen Alpha lingo non-stop. Still, out of all the clues, tone might be the most reliably... well, reliable one

4. Check for disclaimers or qualifications. This is probably the most technical tip. AI would normally add sort of a disclaimer at the end of a text. This is because AI is meant to display the “whole truth”. For example, say you write an argumentative essay about rabies and how it affects human health. The entire purpose of the essay is to antagonise rabies in order to fulfil criteria. However, at the end of the essay, an AI would say something like: “It is important to remember that not all rabid dogs are a threat to human health.” This goes completely against what the writer is trying to achieve in the essay, and therefore is a surefire way of detecting AI when students are writing argumentative essays.

5. Check for weird and out of place errors. AI would never make a simple mistake such as messing up “a” or “an” that most students would make. Instead, it would make factual errors, known as hallucinations. AI does this because AI is fed the entire internet in order to learn, however the internet is full of misinformation, leading the AI to making some very out of the place errors (the most famous example is “Bruckner was a famous painter in the 17th century.” However, Bruckner was a famous conductor and organist, NOT a painter. This actually remained something that AI would say for a very long time, until it eventually got corrected.)

Even after reading this article, please remember there is no surefire way to check for AI (that’s why I said 99.99%). But, with practice, you will be able to separate AI generated text from human text without a clunky website full of ads. In fact, one of the 5 paragraphs above was actually generated by an AI. Good luck finding it. (I especially challenge Mr Coe to do so).