A Grade 7 class in Ontario is asked why a city gets hotter in summer. One student returns with a fluent AI-generated answer. Another arrives with a hand-drawn diagram, a local temperature chart and three questions about tree cover. Both have words on the page, but only one gives the teacher convincing evidence of a student's reasoning. The difference is not whether a chatbot was used. It is whether the learner can explain, test and defend an idea independently.
Across Canada, AI literacy is becoming a practical teaching challenge rather than a specialist technology topic. Students encounter automated recommendations, synthetic images, search summaries, translations and conversational AI long before they learn how those systems work. The goal is not simply to write a better prompt. It is to recognize when AI is present, judge what it can and cannot establish, protect people affected by its use and remain responsible for one's own learning.
This guide offers ten adaptable AI literacy activities for Canadian classrooms, from elementary to secondary school. Most need paper, markers and teacher-prepared samples rather than live chatbot access. Every activity includes a learning purpose, a realistic classroom routine and an observable product that helps teachers distinguish polished output from real understanding. Teachers should review their provincial or territorial curriculum, school-board rules, accessibility obligations and approved-tool arrangements before using an AI service with students.
What counts as AI literacy in a Canadian classroom?
AI literacy combines knowledge, skills and attitudes: knowing that AI systems are constructed by people, recognizing pattern-based outputs and their limits, assessing evidence, understanding bias and privacy, and using or declining technology thoughtfully. The OECD and European Commission's 2026 AI Literacy Framework presents this as preparation for responsible participation in a world shaped by AI. The UNESCO AI Competency Framework for Students addresses human-centred thinking, ethics, applications and system design. Neither framework requires every lesson to involve a chatbot.
For Canadian classroom implementation, MediaSmarts' Use, Understand & Engage framework is especially relevant. It connects digital-media literacy skills such as finding and verifying, privacy and security, representation, ethics, and making or remixing to learning resources for Canada's provinces and territories. The aim is to translate these ideas into assessable classroom actions rather than a stand-alone awareness assembly.
A productive rule for every activity is think first, test next, explain last. Ask students to make an initial judgment before seeing an automated answer; use a source, demonstration or discussion to test that judgment; and finish with a short independent explanation. This makes learning visible even in classrooms with limited devices.
Ten ready-to-adapt AI literacy activities
1. Map the invisible humans behind an AI answer
Best for: Grades 4–8 · Time: 20–25 minutes · Materials: chart paper and one teacher-written chatbot-style answer.
Display an answer to a familiar question, such as why a Canadian lake freezes at the surface first. Ask students to draw a map of everyone and everything that may have shaped the response: people who gathered or authored information, system designers, reviewers, the company operating the model, the teacher choosing the question and the student interpreting the result. Add arrows for the training material, prompt and generated output.
Now circle the places where mistakes, exclusions or values could enter. Students may identify an incomplete collection of examples, a misleading prompt or a confident statement that lacks a source. Clarify that an AI-generated sentence is not an independent expert witness, even when it sounds knowledgeable.
Coaching question: “Whose decisions are hidden inside this answer, and whose judgment is still needed?” Evidence of learning: a labelled system map with a brief explanation of two human decisions. For younger students, use a “robot trained to sort objects” card game; MediaSmarts has an Introducing AI lesson for Grades 1–3 that illustrates this idea.
2. The prompt laboratory: Better instructions, or better thinking?
Best for: Grades 6–12 · Time: 25–30 minutes · Materials: three teacher-prepared responses to differently worded prompts.
Present a weak instruction—“Tell me about floods”—and two progressively clearer alternatives asking for audience, location, scope, uncertainties or supporting evidence. Compare the resulting samples without implying that the longest response is the best. Groups annotate what changed, what remains unverifiable and what the learner still needs to understand.
Then remove the AI output. Each student writes the three questions they would investigate about a real local flood-risk issue, choosing whether an AI system would help, distract or add no value. The point is not prompt optimization as a competitive sport. It is deliberate question formation and task selection.
Coaching question: “If the system disappeared tomorrow, would you still know how to investigate this?” Evidence: a question plan plus a justified decision about tool use. An acceptable answer can be “No AI needed.” Do not ask pupils to enter their names, assignments containing private information or school records into a public chatbot.
3. The evidence ladder: From plausible claim to defensible conclusion
Best for: Grades 5–12 · Time: 30–40 minutes · Materials: a teacher-authored paragraph containing three checkable claims and two trustworthy reference sources.
Provide a short statement about an environmental or scientific topic, for example migratory birds or energy consumption, written in the fluent style of an AI answer. Students highlight three factual claims and place each on an evidence ladder: unsupported; supported only by the generated paragraph; supported by a secondary source; or supported by an appropriate original or authoritative source. They search or consult teacher-supplied reference material and record what the sources actually establish.
Explain that two sites copying the same assertion are not necessarily independent confirmation. They should note uncertainties, scope and publication dates when relevant. This is a compact verification exercise, not an attempt to turn pupils into professional fact-checkers in one period.
Coaching question: “Which exact piece of evidence would change your mind?” Evidence: a claim–source–judgment sheet and one revised sentence. For an extended lesson, add a second round of independent source checking and ask students to explain how the strength of evidence changed their conclusion.
4. The data detective: Teach a deliberately unfair classifier
Best for: Grades 5–10 · Time: 30 minutes · Materials: sets of coloured-shape or fictional-animal cards.
Give teams a tiny training set: nearly all “safe for the park” examples are blue triangles, while nearly all “unsafe” examples are red circles. After students infer the rule, reveal new examples that break those patterns. Different teams get different biased or incomplete training sets. The outcomes will vary even though everyone is trying to solve the same problem.
Ask how the examples, labels and choices influenced each team's decisions. Discuss why classroom analogies simplify real machine-learning systems but illuminate a crucial point: if examples are incomplete or skewed, predictions may be unreliable or unfair. Extend the activity to a fictional school library recommendation system, never to ranking real classmates by ability, ethnicity, behaviour or personal characteristics.
Coaching question: “What new examples would make your decision more trustworthy?” Evidence: a revised training set and written explanation of one failure. Do not equate every error with discrimination; distinguish missing examples, poorly chosen labels and unequal effects.
5. Synthetic-media detective: What can an image really prove?
Best for: Grades 6–12 · Time: 25–35 minutes · Materials: licensed or teacher-created photos, AI-generated illustrations and captions with provenance disclosed to the teacher.
Show students several images described as a Canadian community event, a wildlife sighting or a historic scene. Their job is not to guess “AI or real” from strange fingers or visual glitches. Such visual clues are unreliable. Instead, teams trace the claim: who first shared the image, whether it comes with a credible date and location, and whether another independent record confirms the event.
Reveal the provenance after discussion. A genuine photograph may have a false caption, and a synthetic image can be labelled transparently and used ethically for illustration. The deeper lesson is that authenticity, context and truthfulness are related but different questions.
Coaching question: “What does this picture establish—and what does the caption merely claim?” Evidence: a three-column table labelled image, claim and independent verification. Adapt with printouts when internet access is unreliable, and avoid emotionally distressing or deceptive images of real students.
6. Translation with responsibility: Whose meaning gets lost?
Best for: Grades 7–12 · Time: 30–35 minutes · Materials: short teacher-approved passages, human-reviewed translations and vocabulary notes.
In multilingual Canadian classrooms, translation tools can assist comprehension but can also flatten context. Use short public-domain or teacher-written passages with idioms or subject-specific vocabulary in English and French, or another language represented by qualified classroom expertise. Students compare a literal machine-style translation with one reviewed by a fluent speaker and annotate differences in register, implied meaning and cultural context.
This is not an exercise in asking a multilingual child to represent an entire language community. In particular, avoid appropriating or “correcting” Indigenous-language material without appropriate permission, expertise and community guidance. A small bilingual science instruction can work well without such risks.
Coaching question: “Which meaning would a reader miss if they accepted the smoothest translation?” Evidence: two annotated differences and a revised version with a justification. Students should discover that fluent wording can be useful without being automatically faithful.
7. The privacy decision room: Should this information go into a prompt?
Best for: Grades 4–12 · Time: 20–30 minutes · Materials: fictional prompt cards and three category signs: permissible, ask an adult/teacher, do not share.
Read out examples: “Explain the water cycle”; “Summarize my classmate's support plan”; “Write feedback about a student's behaviour using their full name”; “Generate quiz questions from this public science text.” In groups, pupils sort cards and defend the level of risk. The teacher then explains the difference between public facts, confidential student information, identifiable data and material authorized for classroom use.
The Office of the Privacy Commissioner of Canada's guidance calls for safeguards, proportionality, transparency and particular care where children are concerned. The applicable school and provincial privacy rules still govern actual classroom decisions; this card sort is not legal advice.
Coaching question: “Can we reach the learning goal without sharing this information at all?” Evidence: a revised, privacy-preserving prompt and an explanation. Prefer invented examples rather than real student records, and do not require students to create an AI account for this activity.
8. Build a recommendation algorithm with paper cards
Best for: Grades 4–9 · Time: 30 minutes · Materials: fictional book cards, preference tokens and ranking rules.
Each group acts as a recommendation system for a school library. First, they rank ten invented books using a visible rule such as preferred genre. Next, introduce a different goal: include unfamiliar genres or promote books by lesser-known authors. Students watch how the ordering changes and debate which objective they would choose for a school that values discovery as well as convenience.
Draw a boundary between the analogy and a real recommendation platform: commercial systems use far more complicated methods and objectives. Nonetheless, learners can now ask useful questions about their music feeds, video suggestions and news recommendations. Who selected the ranking goal? What might never appear? How might their own behaviour affect future recommendations?
Coaching question: “Which recommendation system would help you learn rather than merely repeat what you already like?” Evidence: two rankings, a transparent rule and a reflection on trade-offs.
9. The counterargument studio: Think before you generate
Best for: Grades 8–12 · Time: 35–45 minutes · Materials: one debatable, age-appropriate school policy question and a simple argument map.
Select a topic such as whether a school should introduce an optional phone-free study block. Students individually write an initial position supported by reasons and evidence. Only then do groups receive a teacher-prepared set of counterarguments—some strong, some weak, some containing assumptions or missing evidence. They identify the strongest objection to their own view and revise their claim.
If school-approved AI access exists, the teacher may demonstrate a chatbot producing alternative arguments; otherwise the teacher-created set is sufficient. The output must never replace the initial independent position. Make space for students to disagree respectfully, and distinguish an argument's clarity from its factual truth.
Coaching question: “Which opposing point improved your thinking, even if you did not change your conclusion?” Evidence: an initial and revised argument map with a short account of the revision. The valuable product is intellectual flexibility, not rhetorical volume.
10. The no-AI teach-back: Can the learner explain it alone?
Best for: Grades 4–12 · Time: 25–40 minutes · Materials: a recently studied idea, exit ticket and partner feedback sheet.
End the unit with a concept that students may have investigated using approved tools: the greenhouse effect, a mathematical procedure, a Canadian historical development or the structure of an argument. For three minutes, pupils close all tools and create a drawing, explanation or example from memory. Next they teach the concept to a partner who is instructed to ask “why,” “how do you know?” and “what would change if…?” Finally each learner returns to evidence and corrects one weakness.
This activity is intentionally not an anti-AI punishment. It verifies that assistance has served learning rather than disguised its absence. Students who need accessibility supports may use those supports; the assessment target is conceptual understanding, not a particular handwriting or speech format.
Coaching question: “What can you now do independently that you could not do before?” Evidence: a self-generated explanation, follow-up response and revision note. Use this output as a baseline for the next teaching decision, not as proof that AI has caused or prevented learning.
A five-lesson implementation sequence
Ten activities do not need ten separate weeks. A school could combine them into a modest sequence, keeping the last portion of each period for reflection. Adjust the order, grade level, available resources and curriculum alignment locally.
| Lesson | Activities | Core question | Student evidence |
|---|---|---|---|
| 1: Understand | 1 and 8 | Who designed this system, and for what goal? | System map and ranking explanation |
| 2: Question | 2 and 4 | How do instructions and examples shape an output? | Question plan and revised data set |
| 3: Verify | 3 and 5 | What establishes a trustworthy claim? | Evidence sheet and image provenance check |
| 4: Act responsibly | 6 and 7 | Whose meaning and privacy must we protect? | Translation notes and safe-prompt decision |
| 5: Demonstrate learning | 9 and 10 | Can I revise and explain an idea independently? | Argument revision and teach-back |
At the beginning of the unit, ask students what they already think AI is good for, where it may mislead them and which decisions should remain human. At the end, repeat those questions. Compare the quality and nuance of explanations rather than counting how many AI terms students can recite.
A classroom assessment rubric that does not reward polished AI prose
Assess a student's judgment through four observable dimensions. A simple three-level rubric can work across subjects: emerging means the learner identifies a concern but cannot yet explain the reasoning; developing means the learner offers a relevant explanation with partial evidence; independent means the learner tests the idea, recognizes limits and justifies a choice.
| Dimension | What the teacher looks for |
|---|---|
| Understanding | Correctly identifies the role and limitations of an AI system or analogy |
| Verification | Distinguishes claims from evidence and checks an appropriate outside source |
| Responsibility | Identifies privacy, fairness, authorship or community concerns |
| Agency | Explains, revises or transfers learning without relying on a generated answer |
Use feedback such as “Your claim is clear; what independent evidence would support it?” rather than “Make your prompt more sophisticated.” This keeps teacher judgment focused on learning rather than technical display.
What if students cannot use ChatGPT or any generative AI tool?
They can still learn AI literacy. Use printed examples, teacher-generated mock responses, physical sorting games, public diagrams and whole-class demonstrations when permitted. Indeed, unplugged lessons often make invisible assumptions easier to discuss. They also reduce the risk of excluding students because they lack home access, an eligible account, a device or the confidence to use English-language interfaces.
Schools should check their own requirements for parental permission, age eligibility, accessibility, data protection, approved products and record retention. Never upload students' identifiable assignments, special education information or private classroom discussions to a third-party model merely to make an activity more interactive. When assessing a tool, ask whether its use is necessary for the learning objective.
In an English-language Canadian classroom, local context also matters. A rural or northern school, a French immersion class and an urban secondary school may have different connectivity, language needs and community priorities. The questions are transferable; a single assumed national school-board AI policy is not.
The teacher-as-coach approach: Keep the thinking with the learner
An AI literacy lesson can become an impressive show-and-tell that changes little. The decisive shift occurs when the teacher asks for an initial hypothesis, creates a productive challenge, watches how students respond, and gives feedback that helps them reason again. That is the difference between merely demonstrating a tool and coaching the learning process.
Within Learning Coach's teacher-as-coach perspective, the learner remains the person who interprets information, develops a meaningful question, attempts an explanation and owns the decision. The teacher offers structure and feedback without confusing an instant response with durable understanding. A child who produces a flawless AI-assisted paragraph may need more coaching than a child whose imperfect first attempt exposes an important misconception.
Educators interested in building this teaching stance can explore the Learning Coach training pathway. The programme should be reviewed directly for current admission conditions, lesson availability and certification details; mentioning the pathway here does not imply that every listed specialization is presently open or complete.
For related thinking about genuine problem-solving, readers can also consult the Learning Coach books catalogue and look for resources on learning autonomy and problem-solving. Individual book landing pages should be checked before linking them in a published version.
Final thought: The output is not the learning
A student who knows ten clever prompts may still believe every confident answer. Another student who has never opened a chatbot may know how to question a recommendation, challenge a claim, recognize a privacy risk and seek better evidence. Which student is more AI literate?
Canadian classrooms need both technological understanding and independent intellectual habits. The ten activities here turn AI literacy from a slogan into visible classroom behaviour: mapping choices, testing claims, finding missing perspectives, protecting information, revising arguments and teaching ideas back. The test of success is not whether students can produce more words with a machine. It is whether they can make better judgments when the machine is silent.
Resources and references
MediaSmarts — Use, Understand & Engage: A Digital Media Literacy Framework for Canadian Schools. Canadian K–12 digital-media literacy framework and province/territory-linked teaching resources.
MediaSmarts — Learning with AI. Grades 4–6 lesson examining verification, ethics and the student as an active learner.
OECD and European Commission — Empowering Learners for the Age of AI (2026). AI literacy framework for primary and secondary education.
UNESCO — AI Competency Framework for Students (2024). Competencies across human-centred approaches, ethics, application and design.
Office of the Privacy Commissioner of Canada — Principles for Responsible, Trustworthy and Privacy-Protective Generative AI. Privacy safeguards and responsibilities for organizations using generative AI.
