What AI detector does canvas use?
Canvas has no AI detector of its own. It's a learning management system that displays results from whatever third-party tool your school licensed — most often Turnitin, sometimes Copyleaks, Unicheck, or SafeAssign. Whether AI detection runs at all is set per assignment by your instructor, which is why it varies course to course.
Why — the first-principles explanation
Canvas is plumbing. Instructure built it to move assignments, submissions, and grades between students and teachers — not to analyze prose. When you submit a paper, Canvas checks whether the instructor selected a plagiarism-review provider for that specific assignment. If yes, it hands your file to that vendor's service, waits, and displays whatever number comes back inside SpeedGrader. Canvas does not compute the AI score and does not interpret it. It's a window, not a detector.
This architecture explains the confusion people run into. The same student, in the same school, in the same week, can have one assignment scanned and another not — because the setting lives on the assignment, not the platform. There's no "Canvas AI policy" to learn. There's an instructor checkbox.
It also means every accuracy question routes to the vendor, not to Canvas. In practice that vendor is usually Turnitin, whose integration is the most common pairing in U.S. higher education. Turnitin's own documentation sets the terms you should know: it needs at least 300 words of prose, it puts an asterisk on any score from 0 to 20% because false positives cluster there, it reports AI-generated and AI-paraphrased text as separate categories in different highlight colors, and it states outright that the score may misidentify text and must not be the sole basis for adverse action against a student.
And the vendor's limits are real limits, not fine print. Stanford researchers found seven commercial detectors flagged 61.22% of TOEFL essays by non-native English speakers as AI-written while classifying U.S. eighth graders near-perfectly. Vanderbilt disabled Turnitin's AI detector entirely on August 16, 2023, reasoning that a 1% false positive rate against their 75,000 annual submissions meant roughly 750 wrongly flagged papers. So the honest answer to "what does Canvas use" is: whatever your school bought, if your instructor switched it on — and that tool is less certain than its percentage makes it look.
An example that makes it click
Think of Canvas as a TV. You ask "what channel does the TV make?" and the answer is: none. It doesn't make anything. It shows whatever channel is plugged into it, and someone else decides which one — or whether to plug anything in at all.
So when your friend says "Canvas caught my roommate," what actually happened is: the school subscribed to Turnitin, the instructor ticked a box on that one assignment, Turnitin ran the scan, and Canvas displayed the result. Change any of those three things and nothing happens at all. Same TV, no picture.
How to do it
- Open the assignment page in Canvas and look for a plagiarism-review provider name — Turnitin, Copyleaks, Unicheck, or SafeAssign — usually shown near the submission details or in the rubric area.
- After submitting, check whether a similarity or AI indicator appears next to your submission. No indicator usually means no provider was enabled for that assignment.
- Read the course syllabus for the AI policy. The tool tells you nothing about what's permitted; the syllabus does, and the two are frequently out of sync.
- Instructors: the setting is per-assignment, under the submission settings where you choose a plagiarism review service. Enabling it on one assignment does nothing to the others.
- Instructors: treat the number as triage. Turnitin requires 300+ words and asterisks anything from 0 to 20%, and its own guidance forbids using the score alone to act against a student.
- Students who are flagged: ask which vendor produced the score, the exact percentage, the word count, and the highlighted passages — then bring your Google Docs or Word version history.
Key facts
- Canvas ships no native AI detector; it displays AI and similarity results produced by third-party plagiarism-review providers integrated at the assignment level, most commonly Turnitin.
- Turnitin requires at least 300 words of prose and displays an asterisk on AI scores between 0 and 20% because false positives are more frequent in that range.
- Turnitin reports two separate categories in its Submission Breakdown — AI-generated text (highlighted cyan) and AI-generated text that was then AI-paraphrased (highlighted purple).
- Turnitin states its AI writing detection may misidentify human-written, AI-generated, and AI-paraphrased text and should not be used as the sole basis for adverse action against a student.
- Vanderbilt University disabled Turnitin's AI detector on August 16, 2023, calculating that a 1% false positive rate against its 75,000 papers submitted in 2022 would wrongly flag about 750 papers.
- Seven commercial detectors flagged 61.22% of TOEFL essays by non-native English speakers as AI-generated while classifying U.S.-born eighth graders' essays near-perfectly (Patterns, 2023).
▶ The 60-second explainer (script)
What AI detector does Canvas use? None. Canvas doesn't have one. Canvas is a learning management system — plumbing. It moves your assignment from you to your teacher. It doesn't read your prose. What actually happens is this: when your instructor set up that specific assignment, they either did or didn't pick a plagiarism-review provider. If they did, Canvas hands your file to that vendor, waits, and displays the number that comes back inside SpeedGrader. Canvas doesn't compute the score and doesn't interpret it. It's a window. That's why detection seems random — one class scans, the next doesn't, same school, same week. The setting lives on the assignment, not the platform. In practice the vendor is usually Turnitin. Sometimes Copyleaks, Unicheck, or SafeAssign. And every accuracy question belongs to them, not Canvas. Turnitin's own docs are worth knowing: it needs at least three hundred words, it puts an asterisk on any score under twenty percent because false positives pile up there, and it says flat out that the score can't be the sole basis for punishing a student. Stanford found seven detectors flagged sixty-one percent of essays by non-native English speakers as AI. Vanderbilt turned Turnitin's detector off completely — at one percent error across seventy-five thousand papers, that's seven hundred and fifty innocent students a year. So: check the assignment page for a provider name, read the syllabus for the actual policy, and if you get flagged, ask which vendor, what score, how many words — and bring your version history.
What authoritative sources say
People also ask
Can Canvas see what I do in other browser tabs?
No. Canvas can log submission times and, in quizzes, some page-focus events, but it cannot see other tabs or applications. Separate proctoring products can do more, and those are visibly installed and announced — Canvas by itself is not one.
Does every Canvas assignment get scanned for AI?
No. Plagiarism review is enabled per assignment by the instructor. If no provider was selected, no scan runs and no AI data appears at all, which is why the experience differs from course to course.
Does Canvas flag copy-paste?
Not natively. Some other platforms do — NoRedInk's Originality Insights reports pasted-text percentage and time per draft — but a plain Canvas text-entry submission doesn't produce that kind of process data.
My Canvas assignment shows an AI score. Is it proof?
No. It's the vendor's estimate, and Turnitin's own documentation says it may misidentify text and cannot be the sole basis for action. Ask for the word count and highlighted passages, and offer your document's revision history.