As a linguistics professor who is currently teaching in the middle of the generative AI boom, I have been thinking about how we can use AI not just as a tool, but as a collaborator in fostering greater student learning. I have discovered, somewhat accidentally, that the ways AI can be wrong can sometimes produce the most generative teaching and learning moments in the classroom. Like many instructors, I notice that even when AI gets it wrong, the mistake still sounds polished and authoritative.
Instead of looking at this as a problem, I can leverage these errors as launch points for critical thinking, validation of research, and active learning. My students become investigators, editors, and curators of knowledge, not just learning what is right, but why it is right, and how to identify what is wrong. When AI gets things wrong, it creates powerful teachable moments. Giving students an AI-produced answer that contains mistakes pushes them to slow down, test claims, and fix problems. Instead of accepting polished prose as proof, they practice asking: Is this accurate? How do I know?
1. Spot the Error
This strategy strengthens students’ core skills. They practice critical reading by spotting inconsistencies, AI hallucinations, and overconfident claims. They build reasoning skills by explaining why something is wrong and how to fix it, deepen disciplinary knowledge by returning to core concepts, definitions, and evidence, and sharpen information literacy by recognizing that not all sources, especially AI, are authoritative. One strategy I use is a short “Spot the Error” challenge. Students receive an AI-generated answer to a linguistics problem, identify each error, justify their diagnosis with evidence, and produce a corrected version. The same structure works across disciplines. In a history class, students can fact-check timelines and causal claims. In science, they can flag flawed reasoning or misapplied principles. In mathematics, they may find the first wrong step and rebuild the proof or solution.
2. Fact Check the Bot
Sometimes I ask an AI tool to explain a concept, then have students evaluate that answer against our textbook or my lecture materials. They decide which parts are accurate and which are misleading or wrong. In this way, students become critical readers and fact-checkers rather than passive recipients as they learn to weigh AI outputs against vetted academic sources.
3. Rewrite the Response
Another teaching strategy I use in my classes is the “Rewrite the Response” task. In this activity, students receive an incorrect AI-generated answer to a question about a linguistic theory (for example, what are the main approaches to language acquisition). They are then tasked with rewriting it in a way that is not only accurate but also incorporates two additional theoretical perspectives from the literature. This task goes beyond simply spotting errors; it requires students to move from correction to creation. They must demonstrate their ability to identify mistakes and produce a more comprehensive and improved answer. This process encourages engagement with theories, the consideration of multiple viewpoints, and the use of academic writing. It is especially suitable for linguistics, where comparing and analyzing competing theoretical frameworks is often necessary. Moreover, it fosters critical thinking and analysis of theoretical concepts, and encourages constructing richer, more complex arguments supported by empirical evidence.
4. AI vs. Human Reasoning
In this exercise, students receive an AI response about a topic and consider whether they agree or disagree with it. Their assignment is to develop a reasoned argument that either supports or contests the AI response, using a linguist’s eye for evidence and disciplinary frameworks to justify their position. The activity is particularly effective in areas such as sociolinguistics, language acquisition, history, and other fields where argumentation and interpretation are more central. Through this exercise, students develop critical thinking skills while sharpening their ability to articulate their position fairly and clearly.
5. Debate the Bot
In this activity, students read an AI-generated answer to an argumentative prompt, then work in pairs or groups of three to build a case that either supports or challenges it. They must use evidence and appropriate frameworks to justify their position. Here, AI is a conversation partner, not the authority. Students weigh its claims against their own reasoning and theories covered in the course. This approach is especially effective in classes where interpretation and argumentation are central. Instead of searching for errors, we treat the AI response as a starting point for analysis and debate. I’ve observed students using it to articulate clear, well-supported positions.
6. Teaching Academic Verifiability: Cite Check the AI
A new opportunity for learning came when I discovered that AI often cites fabricated references that are formatted perfectly and look credible, but do not actually exist. Rather than eliminating the use of AI altogether, I turned this into a learning opportunity for my students, which I titled “Cite Check the AI.” In this activity, students are given AI-generated responses to questions such as, “What evidence supports or challenges a critical period for second language learning?”
Their first task is to check the citations in the response. Specifically, students must (1) determine whether the references are real and credible, (2) explain in their own words why fabricated references are a problem, if they find any, (3) revise the reference list using reputable scholarly sources they locate themselves, and (4) consider the risks of using AI-generated references in academic writing. To support them in this task, I remind students to use library databases or Google Scholar to verify citations, and I offer guidance on checking DOIs, journal titles, and peer-review status. This strengthens their research skills and reinforces the importance of academic verifiability.
These activities aren’t about “catching AI” making mistakes. They’re about developing core academic skills. AI can be helpful, but it isn’t neutral or flawless, and that’s exactly why these exercises matter. They give students space to practice judgment, question sources, and build a confident scholarly voice so they can participate in academic conversations with genuine ownership. In other words, what appears to be an AI flaw turns into a teaching advantage.
Recent research supports these pedagogical methods. Tzirides et al. (2024) note AI should function as a supportive tool rather than a substitute for human learning. Research shows effective adoption depends on strategies that strengthen critical thinking and creative problem-solving, not displace them. Accordingly, curriculum design should integrate AI thoughtfully while preserving space for unmediated inquiry, creative work, and interpersonal interaction. Moreover, Mollick and Mollick (2023) show that AI’s output can be deployed in multiple classroom workflows. While AI will not replace instructors, well-designed tools can expand instructor capacity, improve learning, and scale evidence-based teaching.
These approaches give students an opportunity to engage with information actively and to reflect as they test, validate, and revise it, rather than receiving it passively. It is also important how we model this as educators. When AI makes a mistake or produces an unreliable example, we can be transparent about it, identify the uncertainty, show how to verify it, and demonstrate how to challenge even polished-sounding claims. That is the habit we want students to develop.
As generative technologies evolve, teaching practices must also adapt by embracing the opportunities, possibilities, and limitations of AI tools rather than resisting the technology as an obstacle to learning. By thoughtfully integrating AI tools into the classroom, we as educators create space for students to collaborate with AI, become more critical and reflective, and avoid using AI as a replacement for their own thinking.
Dr. Safieh Moghaddam (PhD, University of Toronto) is an Associate Professor (Teaching Stream) of Linguistics in the Department of Language Studies at the University of Toronto Scarborough (UTSC). Her research concentrates on sociolinguistics (the interaction of language, culture, gender, and race/ethnicity), linguistic typology and universals (the documentation of minority and endangered languages, as well as the syntax and morphology of endangered and less-studied languages). In the classroom, she focuses on active and cooperative learning and on course design for hybrid (dual delivery) and fully online formats. She has developed and redesigned undergraduate courses at all levels, aligning assessments with clear learning outcomes and creating learner-centered syllabi and activities. Her teaching approach emphasizes consistent, actionable feedback and strong student-instructor relationships to foster engagement, equity, and student success.
References
Tzirides, A. O. O., Zapata, G., Kastania, N. P., Saini, A. K., Castro, V., Ismael, S. A., & Kalantzis, M. (2024). Combining human and artificial intelligence for enhanced AI literacy in higher education. Computers and Education Open, 6, 100184. https://doi.org/10.1016/j.caeo.2024.100184
Mollick, E. R., & Mollick, L. (2023). Using AI to implement effective teaching strategies in classrooms: Five strategies, including prompts. http://dx.doi.org/10.2139/ssrn.4391243
