Avoiding Issues with AI Detectors

Alexandra Parham

        When an assignment calls for instructors to assess students’ writing skills, it is necessary that students and not computers complete the writing. But, in this era of emerging generative AI, how can teachers ensure that students are learning writing standards and demonstrating mastery? How can we best use our limited time and resources to meet instructional goals and the needs of our students rather than searching for computer generated content? Many instructors have been frustrated with the amount of time spent screening student work for AI use that is inappropriate for the goals of an assignment. Since this is such a common frustration, companies, products, and programs have arisen that purport to help teachers identify AI like Turnitin or GPTZero. Although I haven’t made a practice of it, I’ve used these programs myself when I sought to confirm my suspicion that I was reading the work of a computer, not one of my seventh grade students. However, just as students should be judicious about the technologies they use in writing, teachers should be cautious about outsourcing their thinking about and assessment of student writing to these programs.

        According to GPTZero’s website, the company launched its AI detection platform three months after the release of ChatGPT. The company “knew the new challenge that AI could pose to information transparency and academic integrity” (“AI Detector - the Original AI Checker for ChatGPT & More”). Seeing the negative consequences that the technology was already having in classrooms, the company created another product to solve issues with students using generative AI. While there are free versions of AI detectors for teachers to use, there are also more detailed reports available through paid versions of the product. On the other hand, there are free versions of generative AI for students to use and also paid versions that claim to be more realistically human. The result could easily become an arms race between students using more and more advanced generative AI tools and teachers using increasingly sensitive AI detectors, with companies profiting from both sides. Like most educators, I became a teacher because I like kids and I don’t wish to view my students’ work with scrutiny and suspicion.

Currently, the homepage of GPTZero claims to be the world's most accurate AI detector with 99% accuracy. However, when you scan a piece for AI, the results also state “The nature of AI content is changing constantly. These results should not be used to punish students.” The website also boasts about the large number of teachers already using this technology regularly, so even a high level of accuracy is going to generate false reports. False positives have a negative impact on students and so accusations of academic dishonesty should be carefully considered. AI detectors have been shown to have bias against English language learners, students with Autism, and even novice writers. For teachers working in middle school, with English learners, or with emergent writers, all of our students are novices. According to Gegg-Harrison and Shapiro, AI detectors looks for patterns such as generalizations and boosters but “empirical research on student writing  has found that novice writers tend to use these patterns frequently ask well; they need to learn over time how to offer more contextualized and cations claims” (29). AI detectors are more likely to generate a false positive for these students because they may write in a more simplistic and formulaic way as they develop their skills. These students need time and practice to develop their unique voice.

        The consequences of these false reports can vary greatly based on the level of the student and the choices of the academic institution. For older students in high school or college, accusations of plagiarism and academic dishonesty are very serious and can result in failing classes, losses of scholarships, or even exclusion from programs and institutions. For younger students, such as those in middle school or early high school, the consequences may be less dire, but still demoralizing. In my middle school English department, when a student is found to have submitted an assignment that is not their own work, they are given a zero, a behavior referral, and an opportunity to complete the assignment. According to Kevin Roozen in his contribution to Naming What We Know; Threshold Concepts of Writing Studies, writing is “not so much about using a particular set of skills as it is about becoming a particular kind of person, about developing a sense of who we are” (51). Educators do not want to deter students from developing that identity by accusing their writing of being stilted or robotic.They don’t want to imply low expectations by assuming a work is too grammatically perfect to be produced by the student.  Additionally, wrongfully accusing a student of cheating damages the relationship of trust, respect, and support that teachers strive to build with their students.

        Assuming that an instructor does not wish to read and grade assignments generated by computers or engage with the technological arms race of AI generation marketed towards students and AI detectors marketed towards educators and institutions, how can they dissuade students from using AI and ensure academic rigor? The first step is to set clear expectations about what technologies students can and cannot use on a given assignment. For example, Grammarly, once little more than an additional spell and grammar check, has become a lot more sophisticated and can rewrite portions of student work to add more complexity, adjust tone, or clarify the message. If the goal of the assignment is to organize thoughts into clear paragraphs, Grammarly could be a useful tool to help students form cohesive sentences. However, if the instructor is assessing students’ ability to write clearly or vary sentence structure, Grammarly would not be an appropriate tool because too much of that thought process is removed from the student. Now that students have access to a wide variety of AI tools, usually on school issued devices, instructors must be clear about the goals of each assignment, what writing skills are being assessed, and what tools are allowable for that assignment. Of course, assistive technologies such as predictive text or speech to text should always be available to students who require them to meet the goals of the assignment.

Secondly, writing instructors must focus on the writing process rather than the product. In Naming What We Know, contributors Collin Brooke and Allison Carr assert that “when assessment is tied too completely to final products, students are more likely to avoid risking failure for fear of damaging their grades” (63). The goal, then, should be improvement throughout the process and not a perfect final draft. Students at all levels should reach the understanding that writing is a process and that throughout that process they may have successes and failures. Brainstorming, mindmapping, outlining, research notes, rough drafts, and reflections can be used as artifacts of this process. For beginning writers, such as middle schoolers, teachers usually dedicate class time to guiding students through the steps of the writing process and therefore some steps such as brainstorming and outlining may be paper artifacts or even separate assignments. Older students and more proficient writers are often able to choose which parts of the writing process they will employ in the creation of a piece, much of which occurs individually or outside of class time. However, even the most practiced writers do not jump from a concept to a polished and publishable final draft without intervening steps. Some technologies, such as the version history of the document or extensions like Draftback and Revision History can help teachers and students identify these artifacts in a finished work. In short, rather than scouring a finished product for evidence of plagiarism and AI, engage with students throughout their writing process as they strive for improvement. By the time my students are completing a large writing project, I’ve met with each of them several times. If the final product is not a continuation of the work that we’ve been engaged with in class, that’s an indication that it was perhaps not written by the student.

Next, writing classrooms should be social places that emphasize communication. Even when done in isolation, writing is a social act. It’s a way to communicate ideas and feelings to a reader. According to the NEA, “Learning happens, and knowledge is constructed through social engagement and collaboration, making interpersonal interaction between students and educators irreplaceable” (National Education Association). By emphasizing the social aspects of writing such as giving and receiving feedback and collaboration, instructors can incentivize students to engage with writing rather than rely on a machine to make their meaning clear to classmates. In this same vein, instructors should consider allowing more group work and collaborative writing. Many K-12 educators shy away from group work because it can be difficult to manage behaviors and grade student work. However, the conversations that students have when they are generating ideas, making decisions about form, and crafting sentences together are much more rich and nuanced than when they are simply giving feedback on someone else’s nearly finished product. I was skeptical about allowing students to complete writing assignments collaboratively, but after giving my students the option and listening to the rich discussions that emerged, I now give students that option for many assignments.

Finally, students will be less incentivized to use AI if they have adequate time and support to engage with writing tasks. In K-12 classrooms, most of  the writing process should be undertaken in class and not as homework. When choosing writing tasks that help students develop skills, teachers should consider the scope that is necessary for the assignment. For example, if the goal of the assignment is to convey a mood by describing the setting, students who are motivated to write an entire short story may do so, but other students might only write a descriptive paragraph, similar to the box text that sets the scene before a Dungeons and Dragons adventure. Whatever tasks students choose to undertake to demonstrate mastery, they should have access to templates, examples, and outlines. Throughout the process, students should be able to meet with each other and with instructors to discuss their ideas and receive feedback. When students are working on a larger project, instructors will have collaborated with them several times and therefore would be able to recognize text that was suddenly generated by a computer and not the student. Students will be less likely to turn to generative AI when they feel confident that they will be able to complete the writing task and know that their effort, improvement, and originality is valued more than the technically perfect draft a computer can create. Also, teachers will be aware of students’ thought process.

While writing is an important skill and teachers, especially K-12 teachers, are beholden to teach writing, speaking, and listening standards, there are other, nonwritten ways that teachers can assess students’ knowledge. Is a research paper the best way to assess that students’ were  able to locate credible sources and use evidence to support an argument, or would those skills be best showcased through a presentation or a debate? Must poems be written in short, enjambed lines or can they be performed orally with body movements and facial expressions that convey more meaning than mere words? Is writing a short story the best way for students to show their understanding of characterization, or could performing a skit meet the same learning target? Even when writing is the best way to demonstrate understanding of standards, that writing does not need to occur solely on word processing software. According to Carter and Matzke in the article “The More Technology, the Better” from Bad Ideas About Writing, “when technology isn’t mindfully incorporated into the classroom, it can become a distraction that significantly impacts learning” (321). Teachers should provide students with a variety of opportunities to practice communication skills, written and verbal, digital and analog. Furthermore, many students crave these interactions and are tired of the majority of their assignments being pixels on a rectangular screen.

As writing teachers, our aim is to help our students develop as writers and communicators. To do that, they need to practice writing and communicating. Rather than use the very technology that we are asking students to avoid, educators should create a classroom environment that values improvement throughout the writing process, collaboration, and originality.

Works Cited

Adler-Kassner, Linda, and Elizabeth A. Wardle. Naming What We Know: Threshold Concepts of Writing Studies. Logan, UT: Utah State University Press, 2015.

“AI Detector - the Original AI Checker for ChatGPT & More.” GPTZero, 2023, gptzero.me/. Accessed 30 July 2025.

Carter, Genesea M., and Aurora Matzke. “The More Digital Technology the Better.” Bad Ideas about Writing, edited by Cheryl E. Ball and Drew M. Lowe, West Virginia University Libraries Digital Publishing Institute, 2017, pp. 320–24, textbooks.lib.wvu.edu/badideas/badideasaboutwriting-book.pdf.

Gegg-Harrison, Whitney, and Shawna Shapiro. “From Policing to Empowerment .” Rethinking Writing Education in the Age of Generative AI, 1st ed., Routledge, 2025, pp. 26–41.

National Education Association. “VII. Full Statement Text | NEA.” Nea.org, 2025, www.nea.org/resource-library/artificial-intelligence-education/vii-full-statement-text. Accessed 30 July 2025.

Licensing

Copyright Alexandra Parham, 2025. Licensed CC BY 4.0.