When ChatGPT and other generative AI systems debuted, teachers and professors panicked about them becoming the author of the many assignments they would grade from that point on. The technology soon proved itself to be a poor student, frequently making factual errors and churning out robotic-sounding prose, but the possibility that it’s behind written assignments remains. 

In May 2023, an entire class at Texas A&M was accused of using ChatGPT. The proof? Their teacher ran their work through the AI and asked if it had produced the work, which is a dubious method to check for cheating. If you’ve been accused of cheating using AI, and you haven’t, defending yourself can seem like a Kafkaesque undertaking.

Unfortunately, it’s something that students can face at any point in their academic careers, particularly if they are Black, according to a study by Common Sense Media. It found that 20% of Black teens said they were falsely accused of using AI, compared to 7% of white students. So what do you do if you find yourself in this situation?

The first step to fighting the charge is to have a conversation with your instructor about your research and thought process when approaching the assignment. Demonstrating knowledge will go a long way to removing doubt about your mastery of the subject in question. If this is not sufficient for the accusation to be dropped, you will need to amass evidence and submit it, whether electronically or in person. 

1. Show Your Work With Software

Your best bet for proving you didn’t use AI in your writing is to show the process and progress of your writing. If you used Google Docs to write the assignment in question, you’re in the clear. Simply go to File > Version History > See Version History and you will be able to show your writing process. 

If you composed your writing in Microsoft Word and saved your work to OneDrive or SharePoint, you can go to File > History to show how the document changed over time. 

2. Produce Your Research

With the exception of creative writing, most writing assignments typically involve research to complete. If you have hard copies of notes you took, receipts for library materials, or anything of that nature, you should use those as supporting evidence that you did indeed write your assignment yourself. 

If you did your research online, bring up the browser history to back up your argument. To do this in Google Chrome, click on the three dots at the top right of the screen and go to History > History (if there’s anything embarrassing or NSFW in there, click on the box next to the entry and hit delete beforehand.) In Firefox, open the stacked menu icon and go to History > Manage History. For Microsoft Edge, click on the three dots in the top right and go to History.  

If you have any supporting documents, such as notes on your phone or computer, you should show those as well. All the files should have dates and time stamps that you can use as evidence.

3. Compare for Style

Your writing style has certain hallmarks that should be particularly evident to someone who grades your papers regularly. One of the best ways to show you’ve written a particular document is to present your past writing for comparison. Submitting three to five other papers, no matter the subject, should be sufficient to establish your writing voice. 

4. The Argument Against AI Detectors

Finally, you can also help your case by casting doubt on AI itself. There are numerous tools that can be used to tell if text was written by AI, but it’s important to note that these AI detectors are not a perfect or infallible solution to determining whether a student has used AI to generate a given piece of writing.

In response to the question “Do AI detectors work?” OpenAI (the creator of ChatGPT) says, “In short, no, not in our experience.” The company’s own research has found that AI detection tools can “sometimes suggest that human-written content was generated by AI.” You can share this, along with the many studies that back it up. If you’re a non-native English speaker, you can also cite the recent Stanford study showing that AI detectors have a particular bias for declaring your writing the work of AI. 

Should you want to take things a step further, and you know which AI detector was used to establish the claim against your work, you can test it with published, human-written text and see how it does. If it has a tendency to erroneously declare that human-written text was AI-authored, you can use that as further evidence of its faultiness.