Canvas AI Grading: The Real Answer

If you searched this question a year ago the answer was a flat no, and a great many pages still say so. It changed in 2026, and it changed with conditions attached — which is why "does Canvas have AI grading" now deserves a longer answer than yes or no.

We make one of the alternatives, so weigh accordingly. The check is that the section on Instructure's own feature is longer and more specific than the section on ours.

The short answer

Yes — Canvas now ships native AI grading assistance, as part of Instructure's IgniteAI feature set. It appeared in the Canvas release notes dated 18 April 2026 as "IgniteAI Grading Assistance for SpeedGrader". It generates suggested scores and feedback for written assignments against the rubric attached to the assignment, and the instructor reviews and finalises before anything is released.

Three conditions decide whether that sentence is true of your Canvas: it is a paid-tier feature, it is off by default, and it is rubric-driven. Each of those is covered below, because each is a reason an instructor reads an article like this one and then cannot find the button.

The single most common confusion on this query is a different question entirely. "Does Canvas have AI detection?" — whether Canvas spots AI-written work — is a separate topic with a separate answer, and much of what ranks for "does Canvas have AI…" is about detection rather than grading. This piece is about grading.

What Instructure actually shipped

IgniteAI was announced at InstructureCon in July 2025 as an umbrella framework rather than a single product — an extension of the LTI standard to connect partner tools, Model Context Protocol support so partner agents can join, and a set of first-party capabilities. Instructure also announced a global partnership with OpenAI the same day, whose first deliverable was an LLM-enabled assignment type. The IgniteAI platform itself runs on AWS Bedrock.

The pieces that reached general availability since then, in order:

Discussion Insights — evaluates the relevance and quality of student replies in a discussion.
Rubric Generator — drafts a rubric from the assignment description. Requires Enhanced Rubrics to be enabled.
IgniteAI Translations — course content translation.
Question authoring assistance in New Quizzes.
IgniteAI Agent — launched March 2026, an agentic assistant for multi-step Canvas work: generating rubrics, aligning content, reviewing discussions, creating modules, adjusting due dates. Instructure reported it generally available by mid-2026.
IgniteAI Grading Assistance for SpeedGrader — the one this article is about, in the release notes of 18 April 2026.

Grading Assistance works the way you would hope: it reads a written submission, applies the rubric attached to the assignment, and produces a suggested score with feedback. The instructor edits or overrides it inside SpeedGrader and finalises the grade. Instructure's own framing is that educators keep full oversight and final responsibility for every assessment decision, and the feature does not post grades on its own.

The three conditions that decide whether you have it

1. It is on the paid tiers

In April 2026 Instructure repackaged Canvas into three tiers: Canvas Core (foundational LMS plus essential AI), Canvas Plus (AI for teaching and feedback — rubric creation, discussion insights, translation and grading assistance), and Canvas Next (agentic and advanced, including IgniteAI Agent and AI-powered insights).

Grading Assistance sits in Plus and Next, not Core. Instructure gave institutions free access to the advanced AI features on an announced window — through 30 June 2026 in the US and 30 September 2026 worldwide — after which continued access requires the corresponding tier. If your evaluation happened during that window, check what your institution is actually licensed for now rather than what was switched on in the spring. Instructure does not publish tier pricing.

2. It is off by default

Grading Assistance is a course-level feature option, disabled until someone turns it on in the course's Settings → Feature Options. So "our institution has Canvas Plus" and "my course has AI grading" are different statements, and several university IT pages document exactly this gap. If you are looking for the feature and cannot see it, the order to check is: institutional tier, then admin enablement, then the course feature option.

3. It is rubric-driven

Grading Assistance scores against the rubric attached to the assignment. No rubric, no assistance — which is the correct design, and the same constraint every serious tool in this space imposes. It also means the quality of what you get out is decided by the rubric you wrote, a theme we come back to below.

What it does not cover

Every source describes Grading Assistance in terms of written assignment submissions scored against a rubric. We could not confirm on a primary source that it extends to short-answer or essay items inside New Quizzes, so treat that as unanswered rather than as a yes or a no — and ask Instructure directly if quiz marking is your actual pain point.

The other limits follow from what it is. It is Canvas-only, which matters if your institution also runs a Moodle in another faculty. It marks; it does not build the assessment, generate a question bank, import an existing course, or run a tutor for learners — those live in other IgniteAI pieces or nowhere. And it is licensed as part of a tier, so the AI grading question becomes a whole-platform pricing question rather than a line item.

The non-AI baseline you already have

Worth stating, because a good SpeedGrader workflow already removes more drudgery than most people assume, and it is the thing any AI feature is actually competing against.

SpeedGrader shows one submission at a time with grading controls beside it — points, percentage, or the rubric.
Rubrics score directly, provided the rubric is attached to the assignment and marked for use in grading, with per-criterion comments.
The Comment Library saves reusable submission comments, which is the single highest-return non-AI change most markers can make.
DocViewer annotation on uploaded documents, plus audio and video comments.

If you have not built a rubric and a comment library yet, do that before evaluating any AI feature. It costs an afternoon, it makes the marking measurably faster on its own, and it is the prerequisite for every AI option on this page.

The LTI route: what an external tool adds

The second route is an LTI 1.3 tool. Canvas implements the standard thoroughly: Deep Linking, Names and Roles, and Assignment and Grade Services with line items, scores and results — so a tool can create its own gradebook column, bind it to a resource, and post grades into it. Canvas also supports LTI dynamic registration, so an admin registers a tool from Developer Keys by pasting one URL and reviewing the scopes it asks for.

With EduGears AI the marking flow looks like this:

You build a rubric as a grid of criteria by levels, each level carrying a descriptor and a point value. The grid is reusable across quizzes, essays, projects and standalone assessments.
Learners submit through the activity. Typed text, PDFs, images, and audio or video are accepted; project assignments additionally take Word and PowerPoint files, whose text is extracted before marking.
The AI scores every criterion separately and writes a short justification for the level it chose — so what you review is an argument rather than a bare number.
You read it, change what you disagree with, and post. Your edits are labelled as teacher edits, so the record shows who decided what.
The approved score travels to the Canvas gradebook over Assignment and Grade Services, into that activity's own column.

Two honest qualifications on that flow, because "a human always approves" is a claim vendors overstate. On standalone assessments there is no auto-post setting at all — the result waits until an instructor posts it. Project assignments default to draft-and-approve, and there is an opt-in fully automatic mode that is off unless you choose it. Quiz attempts behave differently: objective items score immediately as they always have, and where a rubric is attached to open-ended items in an attempt, that rubric score is combined into the attempt's result rather than parked for review. If you want a hold point on written work, put it in an assessment or a project rather than an essay item inside an auto-marked quiz.

What the LTI route adds beyond marking is the rest of the course: 26 tools in the same registration, including question generation across nine question types with QTI export, course import and export, slides and study guides, and a tutor grounded in your lesson content. And because the integration is the standard rather than a Canvas feature, the identical setup works in Moodle, Blackboard and Brightspace — which is the argument if your institution is not a single-LMS shop.

Whose standard is being applied

Every route on this page depends on a rubric, so it is worth saying plainly what that means. An AI marker without a rubric is producing a model's general impression of a piece of work, and a general impression is precisely the thing you cannot defend to a student, a moderator or an appeals panel.

AI can help you write the grid — Canvas has a Rubric Generator, and in EduGears AI the Rubrics tool can draft a criteria-by-levels grid from your assignment title and brief. In our case the draft opens unsaved in the ordinary rubric editor under a banner asking you to review every criterion, nothing is attached to an activity on its own, and it is not a rubric until you save it. The AI drafts; you own the standard. That is the line worth holding whichever tool you pick.

It also improves the marking. A criterion with a concrete descriptor — "cites at least two sources and explains their relevance" — is scored far more consistently than an abstract one like "quality of research", by a model and by a human marker alike.

Where Gradescope-class tools fit

A third category is worth knowing about, because it is often the right answer and it is not what people expect from the phrase "AI grading".

Gradescope, a Turnitin company, integrates with Canvas over LTI 1.3 with SSO, due-date sync and grade passback. Its automation is deliberately narrow: AI-assisted answer grouping, where similar responses are clustered so one rubric decision marks many at once, and automatic bubble-sheet marking, plus a code autograder that runs the instructor's own tests. None of that assigns a rubric score to an open response on the model's judgement — the AI reads handwriting and clusters, and the instructor decides what each cluster is worth.

For large-enrolment courses that still run exams on paper — STEM in particular — that is frequently a better fit than anything that scores prose, and we would rather say so than pretend otherwise. Other tools in the Canvas ecosystem sit closer to our end of the spectrum, suggesting rubric-aligned scores inside SpeedGrader. We compare the field in more detail on the best AI tools for Moodle and Canvas page, and head to head on EduGears AI vs Gradescope.

Choosing between the three

	Canvas IgniteAI Grading Assistance	An LTI 1.3 AI tool	Gradescope-class tools
What it marks	Written assignment submissions against the attached rubric	Written, short-answer, project and handwritten work against your rubric	Paper, digital and code assignments, question by question
What the AI decides	A suggested score and feedback for the instructor	A per-criterion score with a justification, for the instructor	Which answers are alike — the instructor scores the group
Where it lives	Inside SpeedGrader	In its own activity, posting over AGS	In Gradescope, posting over AGS
Availability	Canvas Plus and Next; course feature option, off by default	Any LTI 1.3 platform; admin registers once	Licensed separately; admin registers over LTI 1.3
Works outside Canvas	No	Yes — Moodle, Blackboard, Brightspace	Yes — several platforms
Best when	You are already on a paid tier and want marking help with no extra procurement	You want assessment build, marking and grade passback in one tool, across more than one LMS	Large cohorts, paper exams, or code — where grouping beats scoring

Rubric-based AI marking in Canvas over LTI 1.3, with a review step before anything posts — and the same setup in Moodle, Blackboard and Brightspace.

Get started free →

Frequently asked questions

Does Canvas have AI grading built in?

Yes, since 2026. IgniteAI Grading Assistance for SpeedGrader appeared in the Canvas release notes dated 18 April 2026. It produces suggested scores and feedback for written assignments against the rubric attached to the assignment, and the instructor reviews and finalises before grades are released. Three conditions apply: it is part of the Canvas Plus and Canvas Next tiers rather than Canvas Core, it is a course-level feature option that is off until someone enables it, and it requires a rubric on the assignment.

Why can't I see AI grading in my SpeedGrader?

Check three things in order. First, your institution's Canvas tier — Grading Assistance is on Plus and Next, not Core, and the free access window Instructure announced ran only to 30 June 2026 in the US and 30 September 2026 worldwide. Second, whether your Canvas admin has enabled the feature for your account. Third, the course itself: it is a course-level feature option under Settings → Feature Options and is off by default. If all three are in place and there is still nothing, confirm the assignment has a rubric attached, because the feature is rubric-driven.

Does Canvas AI grading work on quizzes?

Every source we could verify describes it as marking written assignment submissions against a rubric. Whether it extends to short-answer or essay items inside New Quizzes is not something we could confirm on a primary source, so we are not going to assert it either way — ask Instructure directly if quiz marking is your real problem. Canvas does separately offer AI question authoring assistance in New Quizzes, which is a different feature: it helps write the questions, not mark the answers.

Is AI grading in Canvas free?

Not permanently. Instructure gave institutions free access to the advanced AI features, including Grading Assistance, on an announced window running to 30 June 2026 in the US and 30 September 2026 worldwide, after which continued access requires Canvas Plus or Canvas Next. Instructure does not publish tier pricing, so the cost is whatever your institution negotiates. If your teachers tried it during the free window, confirm what you are licensed for now before building a term around it.

What's the difference between Canvas AI grading and an LTI grading tool?

Scope and portability, mostly. Canvas Grading Assistance lives inside SpeedGrader, marks written assignments against the assignment's rubric, and is licensed as part of a Canvas tier — a small, well-placed feature with nothing extra to procure if you already have the tier. An LTI 1.3 tool runs alongside Canvas and typically carries the rest of the assessment workflow too: building the assessment, question banks, project and handwritten submissions, per-criterion justifications, and content import and export. It also works identically in Moodle, Blackboard and Brightspace, because the integration is the LTI standard rather than a Canvas feature. Many institutions run both.

Will AI grading be accurate enough to trust?

Accuracy is the wrong single question, and the published research is more mixed than most marketing suggests — agreement with human raters is moderate at best and highly dependent on the task, the rubric and the kind of writing, and at least one large study found scores that differed systematically by the writer's first language. The practical mitigations are the same whichever tool you pick: a rubric you wrote, per-criterion scores you can inspect, a human approving the result, and a check of your own by marking one assignment both ways and comparing where the disagreements fall. We wrote a separate piece on what the evidence actually says.

Does Canvas Have AI Grading?