AI Literacy

What Critical AI Literacy Looks Like on a Tuesday in Class

I recently spent two hours with researchers and teacher educators working on critical AI literacy in Canada. The questions were big. The classroom answers kept coming back to one idea.

It was an online panel on critical AI literacy in Canadian schools. The speakers train teachers, study AI in education, and support schools, and for two hours they talked about power, data, sovereignty, refusal, regulation, and who gets left out. The chat filled with teachers and researchers from across the country and beyond.

Critical AI literacy, as the panel used the term, goes well past knowing how to use a chatbot. It means understanding what the tool is actually doing, whose knowledge it carries, who benefits when you use it, and when not to use it at all.

Most of the conversation lived at the level of policy and teacher education, which is where a lot of this work must happen. But as a retired Head of English, I kept asking the question I always ask: what does this look like on a Tuesday, in a Grade 10 class, with thirty kids and a stack of essays? The answer kept coming back to the same idea. See the thinking while it happens.

Critical AI literacy isn't mainly about teaching students to use AI. For a secondary English teacher, I think one of its most practical expressions is designing learning where students can show what they know, what they did, and how they got there.

Why I Stopped Asking "Did You Use AI?"

Before there were any policies about it, I ran a strong student's essay through an AI detector. It came back with an AI score in the high 90s. In the meeting that followed, she and her parents were adamant that she hadn't used AI, and her father pointed out that detectors are notoriously unreliable. He was right. I had a finished essay and a score, and no record of her thinking along the way. I couldn't prove anything, and she couldn't show her work either: she'd thrown out her rough work after handing in the essay. That wasn't her fault. Submitting process work was never part of the assignment, and I had never assessed any of it.

That meeting is where I stopped asking "Did you use AI?" and started asking "How did you use AI?" The second question only works if the assignment is built to answer it: planning you can see, drafts with checkpoints, a paragraph written in class (perhaps by hand), and a short conversation about the work. It's also why I stopped using detectors altogether, for the reasons set out in the piece on detectors.

One panelist made the same point from another direction. There's still a stigma around AI: using it feels like cheating, so people hide how they use it. Honesty about AI grows only when being honest is safe.

Three Ideas from the Session, and What They Look Like in Class

1. Ask What, Not Why

Transparency came up again and again. Then someone in the chat raised a point I haven't stopped thinking about. For some students, an approved AI tool works as an access support, the way text-to-speech or a graphic organizer does for others. If we require students to explain why they used it, we may be asking some of them to disclose a disability.

In class: ask students what they did with the tool and what they kept. Never require a personal reason for using a tool your school has approved. If a student would rather tell you than write it down, a short conversation counts just the same as a written response.

2. Refusal Is a Choice, and So Is Using It

The panel debated whether you can teach AI literacy to people who refuse to use AI. One panelist compared it to choosing between handwriting and a word processor: nobody should have a technology forced on them. Another raised a harder question: does refusing AI come more easily to students who already have a tutor, family help, or other support at home?

In class: design so that both are honest choices. Students can learn how a chatbot builds a sentence by predicting the next word themselves, on paper. They can judge whether a chatbot's answer makes sense by checking it against the text they're studying. None of that requires an account. Then assess the thinking, not whether a tool was used.

3. Some of the Fairest Evidence Happens in the Room

Access came up too. Some students pay for premium AI, and others have just the free version or none at all. The panel's answer was institutional: schools, not families, should provide the tools. That's right, and worth pushing for, but a teacher can't wait for it.

In class: weight more of the grade toward thinking that happens where everyone works with the same resources, along with the accommodations each student is entitled to: planning checkpoints, a paragraph written in class, and a conversation about the work. Unequal access to paid AI can shape what students bring into the room, so don't build a major assessment where a private subscription changes the help available, unless the school provides equal access. That's also the principle behind the assessment model I built into the IGNITE Framework for AI use in English classrooms: 40% process, 40% in-class thinking, and 20% final product.

What Stays Human

Near the end, one panelist argued that students need time to develop their own writing before leaning on tools that produce language for them. Another speaker, who works with secondary English departments, described teachers asking for practical ways to track conversations and check in at different stages, because the finished essay alone no longer tells them enough.

Making thinking visible doesn't solve everything. A student can still use AI privately between checkpoints and rebuild a process that looks authentic. No assessment design removes that possibility. But the choice was never between certainty and a detector score. It's between weak evidence and better evidence. A plan, a draft, a paragraph written in the room, and a short conversation, taken together, tell you far more about a student's thinking than any single finished product.

That's the work in front of us. Not catching students, but designing assignments where the thinking is visible, so that "Did you use AI?" matters far less than "Show me how you got here."

Where to Start

Pick one assignment you'll give this term. Ask whether a chatbot could complete it for full marks, and where in the process you would actually see a student think. The free IGNITE Assignment Audit walks you through eight questions that answer exactly that, and you can use it starting tomorrow.

For more on where this thinking comes from, read What MIT's AI Report Means for a Secondary English Class and What Ontario's Teacher Federations Say About AI, and What It Means for Your Assignments.

Questions Teachers Ask

About This Piece

This is a personal reflection on an online panel about critical AI literacy in Canadian education. Speakers and participants aren't named, and nothing here is quoted. Their ideas are paraphrased as I understood them, and any errors in the telling are mine. It doesn't speak for the panel, its hosts, or anyone who took part.

Developed by George O'Toole: over 35 years in secondary English, Head of Department, Educational Technology Coach, Faculty of Education instructor. Meet George