Can You Design 'ChatGPT-Proof' Assignments?

Ian Hartley


April 24, 2023

main image description

Much like the debate on the efficacy of using grades as learning indicators in the classroom, ideas on how to design ‘ChatGPT proof’ assignments often center around moving to a focus on ‘process’ rather than ‘performance’, as described by John Warner in his 2022 blog post ChatGPT Can’t Kill Anything Worth Preserving (Warner, 2022), and cited by Kristen Palmer at Butler University, in her online article AI In the Classroom (Palmer, 2022). While many educators would agree that this is a noble pursuit and likely one that can lead to greater learning enrichment and a less stressful student experience, the fact is that our current academic system, and indeed our world, are built to process a grade-based system.

While individual schools may trail-blaze a new learning system that abandons grades, students will face the reality of the grading system, in one of its many manifestations, as they enter higher education, or a job after graduation, where performance is often graded by remuneration, awards, or other forms of achievement, depending on their field.

Allowing students to use systems like ChatGPT to complete their assignments does give them an unfair advantage, and from a purely utilitarian viewpoint, leads to an inefficient allocation of human capital. Of course, from a human and social viewpoint, it’s unfair and creates an environment where those who are not cheating are at a disadvantage. While new learning systems are noble and likely valuable to research, they are still too early-stage to provide solutions to the issue of AI cheating on assignments.

A survey conducted between 2022 and 2023 by the Walton Family Foundation, of K12 students from ages 12-17 years old, found that over 30% of students in that age category had used ChatGPT for school (Walton Family Foundation). Bearing in mind, as of the date of writing this blog post, we’re coming up on barely five months with ChatGPT in existence. Anyone familiar with the adopter curve (pictured below), will see that ChatGPT is growing extremely rapidly, and is already capturing the ‘Early Majority’ segment of the market. Growth is not expected to slow down anytime soon.

test alt

A nearer-term alternative to the discussion that our entire grading system must be changed in response to the rise of AI, is a strategy from some instructors who aim to create ‘ChatGPT proof’ assignments, that are worded in such a way that makes it difficult or impossible for students to use ChatGPT, as also suggested by Kristen Palmer at Butler University (Palmer, 2022). In absence of an alternative solution, this is the most realistic and effective solution to this complex problem. At the very least, this approach leads to higher quality content. While many would not agree that the ‘death of High School English’ is a net positive, I’m sure many can agree that having more enriching, non-template assignments is a net positive. Unfortunately, designing ‘ChatGPT proof’ assignments by enriching prompts isn’t necessarily an effective strategy across the board. Yes, creating a more nuanced prompt or asking a specific question about what a character might mean with a choice phrase picked from a text can stump ChatGPT when you ask it to write your entire essay for you, but it won’t prevent ChatGPT from being used to write your essay if you already have an opinion on that phrase. In essence, these measures represent a significant amount of added effort for instructors, but present only a mild hurdle for ChatGPT.

In reality, ChatGPT is rarely being used to write entire essays that are submitted without editing. Instead, it is being used to create extremely robust outlines of content, or draft the vast majority of an essay, with the student revising and tweaking words so it sounds ‘human’. With this approach, it is nearly impossible to detect the use of ChatGPT.

In the face of these super-powerful tools, philosophical questions begin to emerge about the purpose of education. We believe that AI has the potential to be an incredible force for learning and good around the world, but that it must be operated by competent and well-educated humans. We believe that advancements in technology should not come at the cost of our own capabilities as individuals.

Authoriginal is a simple LTI app and Chrome Extension that completely prevents the use of ChatGPT, other AI assistants, and all other forms of plagiarism on assignments. We are releasing a free tier of Authoriginal in order to help schools adapt to the challenges posed by ChatGPT rapidly. If you’re interested in seeing a demo of Authoriginal, please reach out to us, and we would be delighted to meet with you to help you get set up. Feel free to email us at or call us at (818) 641-5936. We look forward to hearing from you!


Palmer, Kristen. 2022. AI in the Classroom.Butler University Center for Academic Technology,

Walton Family Foundation. Teachers and Students Embrace ChatGPT for Education.

Warner, J. (2022, December 11). ChatGPT can't kill anything worth preserving: If an algorithm is the death of high school English, maybe that's an okay thing. The Biblioracle Recommends. Retrieved on January 11, 2023, from

Get a Demo of Authoriginal

Contact us