Skip to main content
Teaching Strategies 9 min read October 2026

Align AI to the Ontario Curriculum, Not Just the Topic

Almost every AI resource claims curriculum alignment. Almost none of them can tell you which expectation it addresses, or which of that expectation's verbs a student actually has to perform. Here is how to read the document and check the output against it.

A printed Ontario curriculum document with expectation codes highlighted on a teacher's desk

The Quick Answer (AEO Summary)

How do you make sure AI use is aligned to the Ontario curriculum? Work from the expectation, not the topic. (1) Read the overall expectation's verbs, because they name the demand: analyse, investigate and demonstrate understanding are three different tasks. (2) Choose an AI format that matches the strand's mode, so an oral communication strand gets a conversation and an inquiry strand gets an investigation. (3) Check the cross-cutting strand that generated content skips, which is strand A in science and AA and A in Grade 9 math. (4) Mark the specific expectation, then store its code on the resource so the coverage can be audited later. A resource that names a unit but never names a code, or that only reaches the verb "understand", is aligned to a topic rather than to the curriculum.

The Problem With "Aligned"

Topic alignment is easy, and it is not the same thing

Ask a model for a Grade 12 biology resource on molecular genetics and you will get something competent, accurate and organised around the topic. It will also, in almost every case, sit in one register: explain the thing, then ask questions about the explanation. That register is real curriculum, but it is only the last of the three verbs that Ontario science expectations keep returning to. The analysis and the investigation are the parts that take a teacher's judgement to design, so they are precisely the parts a content generator leaves out.

The result is a shelf of resources that are aligned to a unit title and misaligned to the expectation code underneath it. Nobody notices until a student cannot do the application question, or until someone asks in June which expectations the course covered. Reading the document properly takes about ten minutes per course and removes most of that risk.

Four Moves

From the strand on the page to a task you can defend

1

Read the overall expectation for its verbs, not its topic

An Ontario overall expectation is a sentence with three or four verbs in it, and the verbs are the assessment. SBI4U B1 asks students to analyse technological applications of enzymes and assess their social and environmental impacts. B2 asks them to investigate structures and functions. B3 asks them to demonstrate an understanding. Those are three different demands on three different days, and a resource that only reaches B3 has covered one third of the strand named on the label.

2

Match the artifact's mode to the strand's mode

ENG4U strand A is oral communication: listen, speak, reflect. A generated worksheet cannot serve A1 no matter how good the questions are, because the strand is about talking. A Socratic chatbot session, a fishbowl prompt or a recorded discussion can. Ask what mode the strand actually requires before you ask what AI can generate, then pick the format that fits. This single step fixes most misalignment.

3

Check the cross-cutting strand that generated resources skip

Science courses open with strand A, Scientific Investigation Skills and Career Exploration. De-streamed Grade 9 math opens with AA, social-emotional learning skills, then A, mathematical thinking and making connections. These strands carry the inquiry, process and reflection work, and they are the first thing missing when a tool generates content from a topic string. A resource that never leaves the content strands is not fully aligned, however accurate it is.

4

Check the specific expectation, then keep the code on the artifact

Overall expectations are too broad to mark against. The specific expectation is where the actual demand lives: MTH1W C3.2 is not "linear relations", it is representing them with concrete materials, tables, graphs and equations, and interpreting a point of intersection in context. Once your artifact matches a specific code, store that code on the resource. In March, when a department head or a coverage audit asks which expectations this unit addressed, you want the answer to be a field, not a memory.

The Verb Table

What each expectation verb requires an AI artifact to contain

This is the check I run on anything generated. Find the verb in the expectation, then look for the thing it demands in the student-facing material. If it is not there, the resource is not aligned, whatever the header says.

Verb in the expectationWhat it asks of studentsWhat the artifact must containHow to check it
Analyse / assessBreak something real into parts and judge it against criteria.A case, product, policy or data set, plus criteria students apply to it. Not a definition list.Can students point to the criteria and defend a judgement, or are they only recalling?
Investigate / inquiryPlan, perform, record, then interpret evidence.A procedure, an observation log, and a place to record results. A simulation can host this; a reading cannot.Is there something students do, and somewhere the evidence lands?
Demonstrate understandingExplain accurately, in their own words, with correct vocabulary.Explanation, model, diagram or comparison with the correct terms and mechanisms.Would a wrong or vague explanation be visible, or does the task let a student copy a definition?
Represent / connectMove between concrete, numerical, graphical and symbolic forms and explain the links.At least two representations of the same relationship, plus a task that requires translating between them.Does the resource show tables, graphs and equations together, or only one form?
Reflect / self-assessName strengths, next steps and the strategies that helped.A prompt anchored in the work just done, with space for a specific answer.Could a student answer it without having done anything, i.e. is it generic?

Two Worked Examples

Same topic, two very different resources

SBI4U D2 and D3Molecular Genetics

D2: investigate, through laboratory activities, the structures of cell components and their functions in the processes of DNA replication, transcription, and translation. D3: demonstrate an understanding of the processes of DNA replication, transcription, and translation, and the role of genes in the synthesis of proteins.

Aligned: An interactive replication lab where students build a complementary strand and predict the product, paired with an explanation task that uses the vocabulary of the strand: template, helicase, leading and lagging strand, semiconservative.

Misaligned: A five-paragraph summary of DNA replication with bolded key terms. It reaches D3 in the shallowest possible way, and it never touches D2's laboratory investigation requirement.

MTH1W C3.2Algebra

C3.2: represent linear relations using concrete materials, tables of values, graphs, and equations, and make connections between the various representations to demonstrate an understanding of rates of change and initial values, and interpret the meaning of their point of intersection in terms of a given context.

Aligned: A two-context comparison problem where students build a table, graph it, derive both equations, and explain what the intersection means in the scenario. The last part is where C3.2 actually lives.

Misaligned: Fifteen "find the slope" questions. Slope is in the strand, but representation connections and interpreting an intersection in context are not optional extras, they are the expectation.

Our DNA replication lab is built to the first pattern rather than the second: students build the strand and predict the product, which is D2's laboratory investigation, then explain the process in the strand's own vocabulary, which is D3.

Signs the Alignment Is Only Cosmetic

Five patterns I see in resources that name the curriculum

  • The resource names a topic, not an expectation. "Unit 3: Homeostasis" is a chapter heading. It tells you nothing about whether E1, E2 or E3 was addressed, or which of analyse, investigate and demonstrate understanding the task actually asks for.

  • Every task ends in recall. Definition matches, fill-in-the-blank vocabulary and true or false items reach the third verb of most strands and skip the first two. Accurate, and still misaligned.

  • The strand A of the course never appears. No investigation, no career connection, no reflection on learning. Generated resources default to content, and the process strands are the collateral.

  • Codes from a different course appear. Expectation codes are course specific. SBI4U expectations do not apply to SCH4U or to SNC1W, and a resource that reuses them across courses was never checked against the document.

  • The reading level is the only differentiation. Two versions at different reading levels is not the same as addressing the different learners in a destreamed class. It fixes vocabulary and leaves the mathematical or scientific demand untouched.

What Alignment Does Not Mean

Three things this is not

It is not a requirement to cover every expectation with an AI resource. The document says what students must learn, not how the material must be produced. Some of the best-aligned work in a course involves no AI at all, and a lesson that reaches one expectation properly beats a bundle that gestures at nine.

It is not a substitute for professional judgement. The code tells you what the document requires; only you can say whether this class, in this week, with this group of students, is served by it. And it is not a compliance exercise for its own sake. The reason to check the verb is that verbs are where students fail: an accurate description of metabolism does not help a student who was asked to evaluate a technological application of it.

Print This — Alignment Check

Six questions before an AI resource goes in front of students

  • The resource names one or more specific expectation codes, not just a unit title

  • The verbs in those expectations are visible in the student task

  • At least one task requires students to analyse, investigate or connect, not only recall

  • The course's process strand (A, or AA and A in MTH1W) is represented somewhere

  • Representations or laboratory work required by the expectation are present, not just described

  • The code stays attached to the resource so the coverage is auditable later

Questions teachers ask

Which Ontario curriculum document should I align AI resources to?

The current Ontario curriculum for the specific course code you teach, published on dcp.edu.gov.on.ca. Each course has its own strands and its own expectation codes, so there is no general Ontario alignment: an SBI4U code belongs to SBI4U only, and the Grade 9 de-streamed courses (MTH1W and SNC1W) have their own codes that should not be swapped with the older academic or applied courses.

What is the difference between an overall and a specific expectation?

An overall expectation states the broad learning for a strand in one sentence, usually with several verbs, such as analyse, investigate and demonstrate understanding. A specific expectation narrows that into something you can actually teach and assess in a lesson, such as representing linear relations in four ways and interpreting a point of intersection in context. Plan and mark against specific expectations; report against overall ones.

Can AI generate a resource that genuinely meets an investigation expectation?

It can build the container: a procedure, a data table, a simulation or a set of prompts that make students do the investigating. What it cannot do is supply the evidence students produce. If nothing in the resource records what students observed, the investigation expectation is being described rather than met.

How do I show curriculum coverage in a department audit?

Attach the expectation code to the resource when you create it, not afterwards. A record with the code stored on it can be filtered and counted per expectation; a folder of documents with unit titles has to be re-read and re-mapped by hand, usually in June, usually by whoever inherited the course.

Does aligning to expectations make an AI resource good?

No. Alignment is necessary and not sufficient. A resource can name the right code and still be badly pitched, badly scaffolded or badly written. Run the alignment check here first, then the broader quality check for accuracy, grade level and pedagogical structure.

Continue Reading

Pick the Expectation First

The curriculum picker in the generators stores the expectation code on the resource, so the alignment you chose today is still visible when someone asks in March.