Can Canvas Detect ChatGPT Through Turnitin?

Canvas cannot detect ChatGPT. It has no AI detector, no writing analysis and no AI score anywhere in its interface. What can detect ChatGPT is Turnitin, and only when your school licensed AI writing detection, an administrator left it enabled, and your instructor ticked Turnitin on that specific assignment. Miss any one of those and the submission is never scored at all.

The confusion is understandable, because the report opens inside Canvas and the percentage shows up in the Canvas Gradebook. From a student's seat it looks like Canvas did the work. It did not. Canvas moved the file.

Does Canvas use Turnitin?

Only if your institution bought Turnitin and connected it. There is no default. Turnitin publishes three separate Canvas integrations and documents them in its own guides: LTI 1.3, which it recommends for new setups, the older LTI 1.1, and the Canvas Plagiarism Framework that runs inside an ordinary Canvas assignment.

The distinction matters more than it sounds. With either LTI integration, your instructor builds a Turnitin external-tool assignment, and Turnitin runs its own submission inbox alongside Canvas. With the Plagiarism Framework, the assignment looks completely normal to you. Turnitin describes that workflow as designed "to provide a similarity report and spelling and grammar suggestions only", with all grading handled in Canvas SpeedGrader.

Instructors can run both in the same course, using the LTI for one assignment and the Plagiarism Framework for another. So the fact that your midterm essay went through Turnitin tells you nothing whatsoever about your final paper.

Does Canvas automatically use Turnitin?

No, and this is the detail that decides the answer for most people. In the Plagiarism Framework workflow, Turnitin has to be switched on per assignment. Turnitin's guide spells out the sequence: the submission type "needs to be set to Online Submission and either Text Entry or File Uploads (or both) need to be selected", and only when those are chosen does "the option to enable Turnitin for the assignment show up in the Plagiarism Review box".

Which means an instructor building an assignment has to notice that box and tick it. Plenty do not. An assignment where nobody ticked it never reaches Turnitin, gets no similarity check, and gets no AI reading, at a university that pays for all of it.

Stack that on the three conditions Turnitin applies everywhere, and you get four independent gates between your Canvas upload and an AI percentage. We lay all four out side by side, with the sources, on our Canvas AI detector page.

What Turnitin actually detects, if it runs

Where all four conditions hold, Turnitin scores the finished text for the likelihood that a language model produced it. Its published model list is long and current, covering the GPT-5 family and GPT-4o, Claude Sonnet and Opus, Gemini, Grok, LLaMA, Mistral, DeepSeek and others, plus "tools based on these LLMs as well". ChatGPT is squarely inside that.

It is worth being precise about what it works from, because a lot of anxiety comes from imagining more. Turnitin analyzes the submitted document. It has no view of your browser, no view of your ChatGPT account, no access to your prompts and no record of your chat history. If a paper you wrote yourself scores badly, that is the model reading your prose, not evidence of anything it found elsewhere.

Turnitin also checks for rewriting. It says its capabilities include "detection of likely AI-generated content that may have been modified using a word spinner/AI paraphrasing or bypassing tool to evade detection", that this runs automatically on every submission where AI detection is enabled, and that it is folded into the same single percentage. It refuses to publish which tools it catches, on the grounds that a list "would make it easier for students to evade our system".

What never gets checked, even with Turnitin running

A large share of what you hand in through Canvas is not eligible for an AI reading in the first place, and the reasons are structural rather than lenient.

The discussion board row catches people out, because weekly posts are exactly the writing most likely to have been drafted by a model. They escape on two counts at once: the wrong object type in Canvas, and a word count well under Turnitin's floor.

Can Canvas detect copy and paste?

Canvas does not flag pasting into a submission box. There is no keystroke recorder in the Canvas text entry field, and no event fires when a finished draft arrives in one piece. What Canvas does keep is metadata: submission timestamps, page views, and for classic quizzes an activity log of events like leaving the quiz page.

Instructure is direct about how much weight that log deserves. Its guide on viewing a student's quiz log says quiz logs "are not intended to validate academic integrity or identify cheating for a quiz", and that the feature "is designed to help you investigate problems that a student may have in the quiz and gain insight into your quiz questions". Instructors who have tested it describe it firing when a cursor drifts to the browser toolbar and staying quiet when a student genuinely switches tabs.

Tools that really do record how a document was written exist, but they live in the editor you drafted in, not in Canvas. That is the distinction worth holding onto: detectors guess from finished text, while provenance features observe the writing as it happens.

Does Canvas detect Grammarly as AI?

Canvas detects nothing, so the question is whether Turnitin does, and the answer depends entirely on how much the tool changed. Turnitin scores the text it receives and has no way of knowing which browser extensions touched it. Fixing commas and tightening a clause leaves your sentence structure intact, and structure is most of what the model reads. Letting a tool rewrite whole paragraphs is a different operation, and it produces the flat, evenly-paced prose that scores worst.

Turnitin names the writing its own detector is most likely to misread, and the list is instructive: "content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas". Note that none of those describe using AI. They describe uniform writing. Which is why students who wrote every word themselves sometimes get flagged, a problem we quantify in the AI detector false positive rate.

What to do about it

You cannot find out whether your course has AI detection enabled. The indicator is hidden from students whether it is switched on or off, so both cases look identical from where you are sitting. That leaves three moves that hold up either way.

Read your academic integrity policy rather than a forum thread about someone else's university, because all four gates are local decisions. Keep a drafting history, since a document that developed over days answers the question a flag actually raises far better than an argument about detector accuracy. And do the reading, which sounds like a platitude until you notice that the students who get into trouble are usually the ones who outsourced the understanding rather than the typing. If your course material is lecture recordings, running one through a tool that turns a video into practice questions gets you further in twenty minutes than a generated draft ever will.

Then write like the submission is being scored, because assuming it is costs you nothing and assuming it is not can cost a great deal. That does not mean chasing a number. It means prose with real variation in sentence length and structure, specifics instead of category words, and arguments that develop rather than restate. Our humanizer rebuilds AI drafting along exactly those lines, and it is a different operation from swapping synonyms, which tends to make a reading worse rather than better. What it will not do is promise you a Turnitin score, because there is no public Turnitin checker to test against, as we explain on our Turnitin AI checker page.

The short version

• Canvas has no AI detector. Turnitin states there is "no AI writing indicator or score embedded directly in the LMS user interface".
• Canvas uses Turnitin only if your school bought it and connected one of the three published integrations. Nothing is present by default.
• Turnitin is not automatic in Canvas. In the Plagiarism Framework it appears as a per-assignment tick box the instructor has to find and use.
• Turnitin lists the GPT-5 family, GPT-4o, Claude, Gemini, Grok, LLaMA, Mistral and DeepSeek among the models it detects, and checks paraphrased AI text in the same pass.
• It works from the submitted document only. It cannot see your browser, your prompts or your ChatGPT account.
• Discussion posts, quiz answers and anything under 300 words of prose are not scored, whatever your school has enabled.
• The percentage visible in the Canvas Gradebook is the similarity score, not an AI score. The AI indicator is instructor-only.
• Instructure says its quiz logs "are not intended to validate academic integrity or identify cheating for a quiz".