By Brett Vogelsinger
No matter where we stand in our experience or feelings about generative AI, as writing teachers we owe it to our students to teach them something about it.
We do not have to be experts in artificial intelligence to teach students how to reason and how to – as A.J. Juliani puts it – use discernment as they apply “the new skill we didn’t see coming.” This can begin young, even before students are permitted to interact with generative AI in their learning.
For twenty years, I was a middle school teacher before moving the high school just a few years ago. My current students have until recently had the major LLMs (large language models) blocked on their 1:1 laptops, and even now, they can only access a paid version of Microsoft CoPilot that is compliant with federal student privacy laws.
Long before this point, it was clear that plenty of my students were employing generative AI in their writing process, and my middle school colleagues saw the same thing in their classrooms. Sometimes, their AI use was intrusive, taking away the student’s voice and ultimately misrepresenting what the student knows and can do.
We call that cheating. Reasoning is an inoculation against cheating. And we can start young.
Teaching reasoning is pertinent to any subject area, broadly transferrable, and something that adults – even those inexperienced in using generative AI themselves – can model through a simple think-aloud during a single class period.
The prefrontal cortex of the adolescent brain, the part of the brain most associated with reasoning and decision-making, continues developing through the mid-20s (“The Teen Brain: 7 Things to Know”).
As teachers, we have already been through that maturation, so whatever new technology arises, we bring biological advantages to the table when we model reasoning. We do this already, daily.
Additionally, the wisdom we’ve won from life lessons before the broad adoption of artificial intelligence gives us depth when we speak to students about it, even when our students may have more experience with generative AI than we do.
AI Reasoning: Two Things You Might Try
So what does it look like to teach students how to reason about AI? Consider these two classroom examples I have used recently, recipes you can alter to fit your situation. They are adapted from lessons in my new book, Artful AI in Writing Instruction: A Human-Centered Approach to Using Artificial Intelligence in Grades 6-12.
The only assumption built into these examples is that you, the instructor, have access to a generative AI tool that has been approved by your institution for faculty use and can project a live demonstration of generative AI use in your classroom.
The lessons will work even in environments where major generative AI tools are blocked on student equipment – in fact, I developed these two lessons under that exact circumstance.
Lesson 1: The Brainstorm
Step 1: Create a brainstorm as a class. This might use a think-pair-share method, a whole class discussion model after a bit of reading. Brainstorm together as you usually would for the lesson you are planning.
Step 2: Share the brainstorm with a generative AI tool approved by your institution. A prompt might start like this: “We are a seventh grade science class studying ____ and we are brainstorming ____. Here’s what we have so far: ____. What are we missing or what do we need to consider as we continue brainstorming?” Don’t forget to include the context sentence at the start of the prompt. This helps maximize the possibility of finding something useful in the output.
Step 3: Evaluate the AI output. Is there anything there that genuinely helps keep the brainstorm going? Is there something that sparks an idea that was neither in the initial brainstorm nor in the AI-generated list? Is another prompt required to gain more focus or clarity from the output?
Step 4: Evaluate the choice. Was the choice to invite AI into this brainstorm a good one? Why or why not? What did it add or take away from the experience? Did the timing of when we invited the AI to be a part of the brainstorm matter? Did we cross boundaries of academic integrity in our use of the generative AI tool?
Lesson 2: Feedback
Step 1: Examine an exemplar for a writing assignment together. Use a document camera or projector, whatever method you typically use when helping students understand what strong writing for a particular assignment looks like. Talk as a class about what is strong about the work, what needs improvement, and what students should tell a conference partner if they were asked for feedback on this piece. (NOTE: It is important for the rest of this lesson to either use work you have created or an exemplar you have permission to share with a generative AI tool).
Step 2: Share the exemplar with a generative AI tool approved by your institution. Elicit feedback from generative AI. You can model good prompt construction, starting like this: “We an eighth grade English class studying ____ and we are looking for feedback on ____. Here’s what we have so far: ____. What should we work on next in this piece of writing?” It may even work well to include a PDF of the rubric for the assignment so that the generative AI can tailor its feedback.
Step 3: Evaluate the output. Is there anything in the output that could help a writer during revision? Where does the feedback overlap with the human feedback we created in our class discussion and where does it differ? Is any feedback especially poor, distracting, or inaccurate? Does a follow-up prompt make the feedback more focused or manageable for a writer to use?
Step 4: Evaluate the choice. What do we gain from human feedback that is impossible to gain from machine feedback on our writing? How does human reading differ from the way an LLM processes text? Does the timing of when we invited artificial intelligence into the the revision process matter? Did the AI output give us anything that made us talk more or dig deeper into our human feedback on this piece of writing?
I hope a few things stand out about these lessons…
- Start with human thinking. The need to develop our ability to think well is not going away, and it may become an even more valuable commodity in a world saturated with artificial intelligence.
- Evaluate and then evaluate again. When we work with AI, we need to do more than evaluate what it creates. We should evaluate whether it was valuable to touch the technology at all. Half of this lesson is evaluative, and that is where the think-aloud demonstration of reasoning happens before we come back to the required course content we were teaching in the first place.
- Open every door. There is no certain conclusion to which any of these questions lead. The lessons are live, raw, experiential, and we can demonstrate real-time reasoning. We can invite student reasoning that might be different than ours or challenge our beliefs and assumptions. These lessons are not designed to point to a single foregone conclusion. Rather, they get us thinking and talking honestly with students about our AI interactions, opening doors together with curiosity.
Even teachers who loathe the very existence of artificial intelligence and consider any interaction with it an affront or intrusion can use these activities. Students deserve not just lectures railing against generative AI’s use, but demonstrations of why it is sometimes NOT helpful or meaningful. Those experiences will stick longer.
So try these lessons, and ruthlessly point out the flaws or absurdities you see in the generative AI interactions and outputs. Measure the time it takes to prompt around unsatisfying outputs. Recognize when AI does make a meaningful contribution. This lets students witness your reasoning in action, not just your opinion.
The real cost of AI
One additional step to consider for any lesson in AI reasoning: How quickly can some research expose the amount of energy and water consumed for the kind of query you just pursued as a class?
Since these impacts are mostly invisible during AI interactions, it’s important that we bring them to students’ attention so they can decide when they think AI use is defensible and when it is not. Like idling a car, frivolous use of artificial intelligence wastes precious resources: students deserve to know this.
Preparing to live with AI
AI is a big, strange, and sometimes terrifying thing, a monster of sorts, that will challenge us individually, educationally, and societally for the foreseeable future.
Sharpening our reasoning powers about when and how to engage with this technology and when and how to abstain from it will serve us and our students well as we navigate whatever the future brings. The discussions you host in your classroom will guide students by modeling what quality reasoning sounds like as we proceed together with curiosity and caution.
References:
Juliani, A. J. “Discernment: The Most Important Skill in an AI World.” 20 Nov 2025. https://www.ajjuliani.com/blog/discernment-the-most-important-skill-in-an-ai-world
Matson, Owen. “The Monstrous Virtue of AI.” Intralation: Culture, Theory, Pedagogy. 27 May 2025. https://intralation-culture-theory-posthuman-pedagogy.ghost.io/is-ai-the-first-true-monster-in-human-history/
“The Teen Brain: 7 Things to Know.” The National Institute of Mental Health. https://www.nimh.nih.gov/health/publications/the-teen-brain-7-things-to-know
Feature image: Unsplash Plus (Tri Wiranto)
Brett Vogelsinger is a high school English teacher in Bucks County, PA with 20 years of experience teaching middle school. He’s an NBCT who writes about teaching, speaks about teaching, and loves to work in the garden.
Brett’s latest book is Artful AI in Writing Instruction: A Human-Centered Approach to Using Artificial Intelligence in Grades 6–12 (Corwin, 2025). He is also the author of Poetry Pauses: Teaching With Poems to Elevate Student Writing in All Genres (Corwin, 2023). You can connect with him at his website. Follow @theVogelman on X-Twitter, Instagram and Bluesky.
Review a free copy of Artful AI for MiddleWeb.
