Can Canvas Detect ChatGPT? What Instructors Actually Use

Canvas itself cannot detect ChatGPT: it has no built-in AI writing detector. When an instructor sees an "AI" score in Canvas, it comes from a separate tool the school has connected, most often Turnitin, whose AI writing report runs on Turnitin-enabled assignments and only on long-form prose (at least 300 words). Turnitin states that its model can be wrong and should not be the sole basis for action against a student, and several universities have switched the feature off. In practice, instructors rely at least as much on conversations, drafts, in-class work and their course's AI policy as on any detector. This guide explains each piece with the vendors' own documentation.

Last verified: October 12, 2026 against the official documentation linked below.

What Canvas does and does not do

Canvas is a learning management system: it hosts your course, assignments and quizzes, and records activity inside them. None of Instructure's instructor guides for the quiz log, the New Quizzes Moderate page, the Access Report or Course Analytics describes any analysis of whether text was written by AI. Similarity checking and AI detection are added through integrations that appear in an assignment's Plagiarism Review settings or as external tools.

What Canvas does record can still be relevant. In a quiz, the Classic Quizzes log adds a timestamped "Stopped viewing the Canvas quiz-taking page" entry whenever the quiz window loses focus, for example when you switch to another tab, and the New Quizzes log shows each event in an attempt along with its timing. The log does not say which site you went to, but a pattern of leaving the page before each answer is visible. Our guides to the Canvas quiz log and everything Canvas can detect during a quiz cover those signals in detail.

How Turnitin's AI writing report works in Canvas

The most common route is the Canvas Plagiarism Framework. Turnitin's Canvas Plagiarism Framework FAQ explains that an instructor creates a Canvas assignment with Online Submission and Text Entry or File Uploads, then turns Turnitin on in the Plagiarism Review box. Reports open from SpeedGrader. Schools can also use Turnitin as an external tool (LTI) instead; both routes attach to assignments, not to quiz questions.

Where the institution has enabled AI writing detection, Turnitin generates an AI Writing Report alongside the familiar Similarity Report. Turnitin's guide "Using the AI Writing Report" describes how it reads:

• A separate percentage. The AI percentage is "different from and independent of the similarity score". It indicates how much of the qualifying text the model believes could be AI-generated, or AI-generated and then altered by a paraphrasing or text-rewriting tool.
• Highlighted sentences. A submission breakdown bar and highlights show which passages the model flagged, so an instructor sees where in the document the score comes from.
• Low scores are hidden. Turnitin found "a higher incidence of false positives when the percentage is between 0 and 19", so any score from 1% to 19% is shown only as an asterisk (*%) with no number and no highlights.
• Strict input requirements. A file must contain at least 300 words of prose in long-form writing, no more than 30,000 words, be under 100 MB, be written in English, Spanish, Japanese or Arabic, and be a .docx, .pdf, .txt or .rtf file.
• Only prose counts. Turnitin says the model "does not reliably detect AI-generated text in the form of non-prose, such as poetry, scripts, or code", nor short-form writing such as bullet points, tables or annotated bibliographies.
• Language differences. Only the English detector checks for AI paraphrasing and rewriting tools; the Spanish and Japanese detectors do not.

Who can see the AI score

The AI indicator is designed for instructors and administrators. When Turnitin launched the feature in April 2023, the University of Melbourne told staff that "neither the detection indicator nor the report will be visible to students" (University of Melbourne). You may be able to see your Similarity Report, depending on the release setting your instructor chose (immediately, after grading, after the due date or never, per the Plagiarism Framework FAQ), but you should not expect to see the AI percentage. Each institution decides whether the AI feature is on at all.

Turnitin's own stated limitations

Students often assume an AI score is a verdict. Turnitin's documentation says the opposite. The AI Writing Report guide states that the model "may not always be accurate (it may misidentify human-written, AI-generated, and AI-paraphrased text), so it should not be used as the sole basis for adverse actions against a student", and that "it takes further scrutiny and human judgment" together with the institution's policies to decide whether misconduct occurred.

Turnitin has also published figures. In March 2023, its Chief Product Officer wrote that the company aimed for "a less than 1% false positive rate" at the document level, acknowledged "there is still a small risk of false positives", and told instructors they "will need to apply your professional judgment" (Turnitin, March 2023). A May 2023 update added more nuance (Turnitin, May 2023):

• The sentence-level false positive rate is about 4%.
• False positives were more frequent in the first and last few sentences of a document.
• The minimum length for analysis was raised from 150 to 300 words.
• "Real-world use is yielding different results from our lab", which led to the asterisk for scores under 20%.

Those figures are Turnitin's own and date from 2023; the model has been updated since, and Turnitin does not publish a current independent accuracy audit in the guides cited here.

Other AI detectors connected to Canvas

Turnitin is not the only option. Canvas's Plagiarism Review setting accepts other providers, and several of them market AI detection alongside similarity checking. They plug in the same way: on assignments, chosen and configured by the institution. If your course uses one, its name usually appears in the assignment details or the submission page, and your school's teaching and learning site normally explains what it checks. The same caution applies to all of them: an AI score is a statistical estimate, and the decision about what happened is made by people, under your institution's policy.

Free online "AI checkers" that students paste text into are a separate thing again. They are not connected to Canvas, your instructor does not see their results, and their scores often disagree with each other, which is one more reason not to treat any single number as decisive.

Why some universities turned AI detection off

AI detection is optional for institutions, and several have chosen not to use it. On August 16, 2023, Vanderbilt University announced it was disabling Turnitin's AI detector "for the foreseeable future" (Vanderbilt Brightspace support). Its reasons: Turnitin offered "no insight into how it works"; even a 1% false positive rate applied to the roughly 75,000 papers Vanderbilt submitted in 2022 would mean around 750 students wrongly flagged; detectors have been found more likely to label non-native English writing as AI; and student data privacy. The University of Alabama's IT center gave a similar reason: "The fallibility of AI-detection tools is a large part of the decision to disable Turnitin's AI detection report" (University of Alabama).

The bias concern comes from research. A 2023 study, "GPT detectors are biased against non-native English writers" (Liang et al., 2023), found that several detectors "consistently misclassify non-native English writing samples as AI-generated" while correctly identifying native-English samples. And OpenAI, the company behind ChatGPT, withdrew its own AI text classifier: as of July 20, 2023 it was "no longer available due to its low rate of accuracy"; in OpenAI's evaluation it identified 26% of AI-written text while mislabelling human text 9% of the time (OpenAI).

Other institutions keep the feature on with guardrails, such as treating the score as a prompt for a conversation rather than evidence. Which approach your school takes is usually published by its teaching and learning centre or academic integrity office.

What instructors use instead of (or alongside) a detector

Because detectors are uncertain, most guidance to faculty focuses on teaching and assessment design. Vanderbilt's guidance, for example, recommends talking with students early about what AI use is allowed, requiring disclosure and citation of permitted use, redesigning assignments, comparing work with a student's earlier writing, checking citations and facts, and raising concerns directly with the student. Common practices include:

On version history: Google's version history help page explains that opening an earlier version shows who updated the file and what they changed, and that only people with edit access can view it. Some instructors ask students to share a document with edit or version access for exactly this reason.

For quizzes specifically, design does most of the work: question banks, randomised answers, one question at a time, time limits and questions that ask you to apply ideas to a specific case from class. Our guide to Canvas quiz settings explains each option. High-stakes exams may also use proctoring tools such as Respondus LockDown Browser or Honorlock.

Institutional AI policies: the rule that actually applies

Whether ChatGPT is allowed in a given assignment is a policy question, not a technical one. Many universities leave the decision to each instructor within an institution-wide academic integrity policy, so the same student may be encouraged to use AI in one course and forbidden in another. Typical syllabus positions fall into three groups:

• Not permitted. Any AI-generated content in graded work is treated like unauthorised assistance.
• Permitted with disclosure. You may use AI for specific tasks (brainstorming, grammar, practice) and must state how, sometimes with a citation in APA or MLA style.
• Encouraged for specific assignments. The assignment itself involves working with AI and critiquing its output.

Read the syllabus and assignment instructions, and if they are silent, ask your instructor in writing before you use any AI tool on graded work. A short email answer protects you far better than any assumption about what software can or cannot notice.

If your work is flagged as AI-written

False positives are documented by Turnitin itself, so a flag is the start of a process, not the end of one. If it happens to you:

1. Stay calm and ask what was flagged. Ask which passages were highlighted and what evidence is being considered besides the score.
2. Show your process. Drafts, notes, outlines, research sources and document version history show how the work came together.
3. Offer to explain the work. Walking through your reasoning in a meeting is often the clearest evidence of authorship.
4. Learn the procedure. Your institution's academic integrity policy explains your rights, including how to respond and appeal. Student advocacy or ombuds offices can help.
5. Point to the documentation. Turnitin's own guide says the score should not be the sole basis for adverse action; citing it is reasonable.

If you are a non-native English speaker, it is fair to mention the research on detector bias when you discuss a flag, alongside your drafts.

Using AI tools to study, within the rules

AI tools are most useful, and least risky, when they help you understand material rather than produce graded work for you: explaining a concept you missed, walking through a worked example, or quizzing you before an exam. Our guide on how to study with AI tools covers practical ways to do that.

Sources

• Turnitin: Using the AI Writing Report
• Turnitin: Canvas Plagiarism Framework FAQ for instructors
• Turnitin: Understanding false positives within our AI writing detection capabilities (March 2023)
• Turnitin: AI writing detection update from Turnitin's Chief Product Officer (May 2023)
• University of Melbourne: Turnitin AI writing detection tool
• Vanderbilt University: Guidance on AI detection and why we're disabling Turnitin's AI detector
• University of Alabama: Turnitin AI writing detection unavailable
• Liang et al. (2023): GPT detectors are biased against non-native English writers
• OpenAI: New AI classifier for indicating AI-written text (withdrawn July 2023)
• Google Docs Help: Find what's changed in a file (version history)
• Instructure: How do I view a quiz log for a student?
• Instructure: How do I use the Moderate page in New Quizzes?

FAQ

Does Canvas have an AI detector?

No. Canvas has no built-in AI writing detection. AI scores come from tools your school connects, most commonly Turnitin, and only on assignments where that tool is enabled.

Can Canvas detect ChatGPT in quiz answers?

Canvas does not analyse quiz answers for AI. Turnitin's AI report needs at least 300 words of prose in a submitted file, so short quiz answers fall outside it. The quiz log does record when you leave the quiz page, and instructors read answers themselves.

Can my professor see my Turnitin AI score?

If your institution has enabled AI detection, instructors can see the AI percentage and highlighted passages for qualifying submissions. Students generally cannot see the AI indicator.

How accurate is Turnitin's AI detection?

Turnitin reported aiming for under 1% false positives per document and about 4% at the sentence level in 2023, hides scores below 20% because they are less reliable, and says the score should never be the sole basis for action against a student.

Can teachers tell if you used ChatGPT without a detector?

Often they look at other evidence: whether you can explain the work, how it compares with your earlier writing, drafts and version history, and whether sources and facts check out. Rules on AI use are set in your syllabus.

What should I do if I am wrongly flagged for AI?

Ask what was flagged, share your drafts and version history, offer to explain your work, and follow your institution's academic integrity procedure, which includes your right to respond.