How Turnitin Works Inside Canvas: Where the AI Score Shows Up and What It Means

Open a Canvas assignment, click into a submission, and somewhere in that interface a Turnitin score is waiting. Students assume Canvas is the thing watching them. It isn’t. Canvas is just the room where another tool’s verdict gets posted, and where that AI number actually lives takes a couple of clicks most people never make.

This guide walks the real plumbing: how Turnitin attaches to Canvas, the two different ways your school might have wired it, where the AI writing score surfaces in each, and why a student can stare at the same screen and never see it.

Key takeaways

• Canvas has no built-in AI detector. When people say “Canvas caught AI,” they mean Turnitin running through Canvas.
• Turnitin connects two ways: the Plagiarism Framework (a normal Canvas assignment, graded in SpeedGrader) and the LTI 1.3 tool (graded in Turnitin’s own Feedback Studio). Where you click for the AI score differs.
• The first number you see in Canvas is usually the Similarity score. The AI writing indicator sits one layer deeper, and only if your school enabled it.
• By default the AI report is instructor-facing. Students typically see a Similarity Report, if anything, not the AI percentage.
• Turnitin’s AI indicator needs at least 300 words of long-form prose to run, and the number is a probability about style, not proof about how a paper was made.

Canvas didn’t detect anything. Turnitin did.

Let’s clear the biggest confusion first. Canvas is a learning management system. It collects assignments, runs the gradebook, hosts discussions. It does not ship with an AI-writing classifier. So “does Canvas detect AI” has a flat answer: not on its own.

What happens is that schools bolt Turnitin onto Canvas. A submission flows out to Turnitin, Turnitin’s models score it, and the results get piped back into the Canvas interface. The judgment is Turnitin’s. Canvas is the messenger.

This matters because every lever that decides whether you even get an AI score lives on the Turnitin side, set by your institution’s admins. Whether AI detection is turned on, what counts, who can see it: none of that is a Canvas setting. So if you’re trying to understand a score, you’re really trying to understand Turnitin’s behavior inside someone else’s configuration.

The two ways Turnitin plugs into Canvas

Here’s the part most explainers skip, and it’s the part that actually changes where you click. Turnitin attaches to Canvas through one of two integrations, and your school picked one.

The Plagiarism Framework

With the Plagiarism Framework, the assignment is a regular Canvas assignment. Students submit the normal way, you grade in SpeedGrader with Canvas’s own tools, and Turnitin works quietly in the background as a plagiarism-and-AI check layered on top.

A practical tell: because it’s a native Canvas assignment, it works inside Canvas’s Student View preview. The similarity result shows up as a colored indicator on the submission, and you open the full Turnitin report from there.

The LTI 1.3 integration

With the LTI 1.3 integration, Turnitin is more of a guest app embedded in Canvas. Faculty who prefer to grade and leave feedback inside Turnitin Feedback Studio lean toward this one. Note that LTI 1.3 submissions don’t appear in SpeedGrader at all; grading and feedback run through Feedback Studio, and the grade passes back to the Canvas gradebook. Because LTI tools aren’t native to Canvas, this version generally won’t run in Canvas’s internal Student View, which trips up instructors trying to test an assignment as a student.

Neither is “better.” They’re built for different habits: one keeps grading in Canvas, the other moves it into Turnitin. But knowing which one you’re in tells you where the AI score will be hiding.

So where does the AI score actually show up?

This is the question that brings most people here, so let’s be exact.

The number you spot first in Canvas is almost always the Similarity score, the colored badge that estimates text matching against other sources. That is not the AI score. People conflate the two constantly, and they measure completely different things. Similarity asks “does this match something that already exists?” The AI indicator asks “does this *read like* a language model wrote it?”

To reach the AI writing percentage, you go one level deeper:

1. Open the submission in SpeedGrader (Plagiarism Framework) or the Turnitin panel (LTI 1.3).
2. Click through to the full Turnitin report / viewer, not just the badge.
3. Find the separate AI writing indicator or tab inside that viewer.

In the report, the AI section reports a percentage of the qualifying text Turnitin predicts was AI-generated, with highlights showing which passages drew the flag. It’s a distinct tab from the similarity matches, with its own number and its own logic.

Two conditions can make that tab empty. First, your institution has to have enabled AI writing detection; it’s a separate switch from plagiarism checking, and plenty of schools run similarity without it. Second, the submission has to clear Turnitin’s bar for long-form prose. Turnitin sets that floor at 300 words of running prose, not lists, code, or block quotes. Miss either, and there’s simply no number to see.

The student blind spot

Here’s a quiet, important asymmetry. The AI writing report is built for instructors and admins. By default, students don’t see it.

A student might be allowed to open a Similarity Report through Canvas, depending on how the assignment was configured, and even feel reassured by a low match percentage. That view generally doesn’t carry the AI indicator. So a learner can submit, glance at a clean similarity number, and have no idea an AI score even exists on the other side of the screen.

If you’re a student: a tidy similarity figure tells you nothing about the AI tab. They’re separate reports with separate visibility rules. It helps to understand how a detector actually scores a passage before you read too much into any single number, since the AI tab is reacting to writing style, not to anything Canvas knows about you.

A mini-scenario: the score that was sitting one click away

Picture an instructor, call him Mr. Alvarez, grading a stack of 700-word lab reflections in Canvas. Each submission shows a green similarity badge, low single digits. He’s about to enter grades when a colleague mentions the AI indicator.

He reopens one report in the Turnitin viewer, clicks past the similarity matches to the AI writing tab, and finds a high percentage he’d never have seen on the badge. A few students cluster high.

Now the right move is restraint, not a referral. The highlights land on smooth, evenly paced paragraphs, which is exactly the kind of writing these tools tend to over-flag. So he opens the document version history in a couple of cases and watches the drafts grow over a few evenings, edits and all. The rest he handles with short conversations. The percentage was real; it was also one click deep, instructor-only, and not proof of anything by itself. That’s the whole shape of AI detection in Canvas in one assignment.

What that number means (and doesn’t)

The percentage is a probability about style, not a count of machine-written sentences and not a record of how the paper was produced. Detectors output likelihoods, and they’re wrong in both directions. OpenAI shut down its own AI-text classifier in July 2023, citing low accuracy. A 2023 Stanford study (Liang et al., published in *Patterns*, with James Zou as senior author) found several detectors disproportionately misclassified writing by non-native English speakers, which the authors link to the lower perplexity, or steadier word choice, common in second-language prose.

Under the hood, these tools reward uneven, surprising writing and penalize smooth, regular prose. If you want the mechanics, we wrote a full piece on the statistics behind how detectors score text. The takeaway for Canvas: the AI tab measures how *regular* the writing looks, never how *honest* the writer is. Use it to decide whether to look closer, paired with drafts and a conversation, not as a one-click verdict. There are more guides on detection and integrity if you’re setting policy.

Frequently asked questions

Does Canvas have its own AI detector?

No. Canvas is a learning management system with no native AI-writing classifier. When someone says “Canvas flagged my essay,” they mean Turnitin, connected to Canvas by their school. The score is generated by Turnitin’s model and surfaced inside Canvas, so it feels like Canvas did it. The rules and the on/off switch all live on the Turnitin side.

Where exactly does the Turnitin AI score show up in Canvas?

Not on the main submission row, where you usually see a Similarity badge. To reach the AI percentage, an instructor opens the submission in SpeedGrader or the submission details, clicks into the full Turnitin report, and finds the separate AI writing tab in Turnitin’s viewer. The path differs slightly between the Plagiarism Framework and LTI 1.3, but the AI indicator always sits a layer deeper than the similarity number.

Can students see the Turnitin AI writing score in Canvas?

Usually not. By default the AI report is instructor-facing, separate from any Similarity Report a student might be allowed to open. Even when students can view their similarity result through Canvas, that view generally doesn’t include the AI percentage. A clean similarity score doesn’t mean there’s no AI flag on the instructor’s side.

Why is there no AI score on my Canvas submission at all?

A few possibilities. Your school may not have enabled AI writing detection, which is a separate toggle. The submission may be under Turnitin’s 300-word minimum, or mostly lists, code, or quotations it doesn’t score. Or it isn’t routed through Turnitin at all. A blank indicator isn’t a pass or a fail. It means the conditions for a score weren’t met.

Is the Turnitin AI score in Canvas proof that I used AI?

No. It’s a probability about how predictable the writing looks, not a record of how the paper was written. Detectors output likelihoods and err in both directions. OpenAI retired its own classifier in July 2023 for low accuracy, and a 2023 Stanford study found several detectors disproportionately flagged non-native English writers. A high score should start a review of drafts and version history, never trigger an automatic penalty.

Read the screen, then go one click deeper

Turnitin’s AI score in Canvas isn’t mysterious once you know it’s Turnitin, not Canvas, and that the number sits behind the similarity badge, usually out of sight to students, only if your school turned it on. Treat it as a prompt to look closer, weighed against process and voice, and keep the final call human.

Curious how your own writing actually reads to one of these models? Check a passage in a detector yourself, starting with something you wrote by hand, and watch how honest prose behaves. If you’re choosing a tier, you can also see what each plan includes.