Direct answer
Teachers on Canvas most commonly use Turnitin's AI writing detection, delivered through the Turnitin integration that institutions plug into Canvas assignments. Canvas itself does not ship a native "AI detector" that flags student prose; the detection layer students actually encounter is Turnitin's, surfaced inside the assignment and SpeedGrader view the instructor already uses [1]. Turnitin's AI writing detection runs as part of the Similarity Report workflow and is available to instructors in institutional accounts, not to students [2].
Does Canvas use Turnitin to detect AI writing, and how does the integration actually work?
Canvas is a learning management system, not a detection engine. When a university licenses Turnitin, the tool is added to Canvas as an external tool or plagiarism framework, and instructors switch it on per assignment through the assignment's external tool settings [1]. Once enabled, a student's submission is routed to Turnitin automatically at the moment of upload, and the resulting report is attached back to the Canvas submission record [1].
That architecture matters because it explains why detection feels invisible to students. There is no separate "AI check" button, no second upload, and no notification that a report was generated. The instructor opens the assignment in SpeedGrader and sees the report alongside the grade field, which is why a flag can appear without the student ever knowing a scan happened [2].
It also explains the inconsistency students complain about. Whether AI detection is even active depends on the instructor's settings and the institution's licence, so two Canvas courses at the same university can behave completely differently [2]. A draft that sails through one unit can be flagged in another, and that is a configuration difference rather than a change in the writing itself.
What do teachers see in the Turnitin AI writing report inside Canvas — and what do the score bands mean?
The instructor sees an overall AI-writing percentage for the submission, plus a sentence-level breakdown that highlights which passages were identified as likely AI-generated [3]. The percentage is a proportion of the analysed text, not a verdict, and Turnitin explicitly frames the report as a signal that supports instructor judgement rather than replacing it [2].
The score bands are where most confusion starts. Turnitin only reports a numeric percentage when its confidence is high enough; when the detected amount falls below the 20% confidence threshold, the report displays an asterisk (*%) instead of a number [3]. Those asterisk readings are low-confidence signals and should not be read as proof of misconduct [3].
For students, the practical takeaway is that a low or asterisked score is not a clean bill of health and a high score is not an automatic accusation. Instructors are trained to read the highlighted passages in context, check whether flagged sentences are quotations, formulaic academic phrasing, or genuinely generated prose, and weigh the report against drafting history [3]. A flagged paragraph inside an otherwise hand-written essay is read very differently from a document that is flagged end to end [2].
Can students check their own draft for the same Turnitin AI score before submitting it on Canvas?
Not through the institutional route. Turnitin's AI detection is an instructor-facing feature, and students generally cannot run it themselves inside a university Canvas account; the instructor decides whether the report is generated at all [4]. Students are instead advised to review their own drafting process, keep evidence of how the work developed, and tighten their citation practice [4].
That gap is exactly why independent pre-submission checking has become common. Because the report an instructor sees is generated from the same detection model, running a draft through an external checker that reproduces Turnitin's AI and similarity output gives the student a preview of the same signal before the deadline [4]. It converts an invisible, irreversible process into something you can inspect and act on while you still have time to revise.
The workflow that works is straightforward: finish the draft, check it against the same detection logic your teacher will see, and then decide whether the flagged passages need rewriting, better sourcing, or simply a clearer record of your own process [4]. Students who check first stop guessing about what Canvas will show and start fixing specific sentences instead.
If you would rather see the report before your instructor does, turnitin0 lets you run your draft through the same Turnitin AI and similarity checks and hands back the downloadable PDFs — so the flag you fix today is the flag that never reaches Canvas.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Do all Canvas courses use Turnitin AI detection?
No. Turnitin is an add-on that institutions license and instructors enable per assignment, so detection is only active where it has been switched on [1]. A course with no Turnitin integration will not produce an AI writing report at all.
What does an asterisk (*%) in the AI writing report mean?
It means the detected AI-generated text fell below Turnitin's 20% confidence threshold, so the report shows *% rather than a number [3]. Those readings are low-confidence signals, not findings of misconduct.
Can teachers see AI detection results in SpeedGrader?
Yes. When Turnitin is enabled on a Canvas assignment, the report is attached to the submission and opens from the instructor's grading view alongside the student's file [1]. Students do not receive an equivalent in-course view [4].
If my AI score is high, does that prove I used AI?
No. Turnitin presents the percentage and highlighted passages as advisory signals for instructor interpretation, and flagged text can include quotations or formulaic academic phrasing [3]. Context and drafting history carry the real weight [2].
How can I see my Turnitin AI score before I submit on Canvas?
Use an independent pre-submission check that reproduces Turnitin's AI and similarity reports, since the institutional feature is instructor-only [4]. That gives you the same signal while you can still revise.