Can Canvas Detect ChatGPT? What It Can and Can’t See

Does Canvas have built-in AI detection?

Instructure’s published quiz-log guide is the closest thing to a native “what was the student doing” record, and it opens with the opposite of a detector. Quiz logs are there to investigate problems a student had and to see how questions performed. “Quiz logs are not intended to validate academic integrity or identify cheating for a quiz.” The note is repeated: do not use them to identify cheating. If the concern is misconduct, Instructure tells instructors to start with the campus office that actually handles it.

An Instructure employee put the design in one place in September 2023: activity logging, including quiz logs, is limited to running Canvas, helping support, and informing users. Interpretation of that data is subjective. Tools built for integrity checks are integrations, not the LMS. That is why a course can feel “watched” and still have no AI score until someone turns an outside product on.

There is a second, quieter confusion with Canvas’s own generative features, such as question authoring. A button that helps an instructor write a quiz item is not a scan of a student’s essay. If your course page never mentions an originality or AI-writing report, do not invent one.

How Turnitin and other integrations work with Canvas

Turnitin’s note on the Canvas changeover says Canvas added Document Processor, an LTI-based way for third-party tools to attach, and that the older Canvas Plagiarism Framework will be retired over time. Turnitin’s claim for the migration is specific: Similarity and AI writing detection continue to run on native Canvas assignments, as they did under the plagiarism framework, and instructors get Similarity and Originality depending on the license. New customers are pointed at Document Processor from the start. None of that is a switch inside every Canvas course in the world. It is present when the school buys it and the assignment uses it.

Turnitin’s AI writing FAQ describes the number those reports show. It is the share of qualifying prose the model judges was likely generated by AI, or generated and then changed by an AI paraphraser or bypasser. Qualifying text means sentences, so bullets and other fragments can sit outside the percentage. Turnitin says it does not determine misconduct. The indicator “should not be used as the sole basis for action or a definitive grading measure.” The same FAQ says the indicator and the report can be seen through the LMS the school uses, including Canvas. It does not say every student can open that report. Ask, if you need to know who sees it in your course.

Other vendors can be connected to an LMS too. This article does not list them, because a connection at one campus is not a feature of Canvas. The way to know is the assignment itself: an originality statement, a Turnitin inbox, or a report link is evidence. A plain file upload with no such statement is not secretly the same thing. How any one percentage should be read, including why two vendors will not match, is covered in are AI detectors accurate.

What quiz logs and activity records actually show

Classic quiz-log auditing is a course opt-in. Once it is on, each attempt stores a start time and a status with a timestamp. The guide’s statuses are: the student viewed a question, answered a question, stopped viewing the Canvas quiz-taking page, and resumed. “Stopped viewing” means the quiz page lost focus. It can be another tab, another app, or a click away from the window. The log does not name the destination. Logs are kept for six months.

Autosave makes some of those rows a bad timeline. For fill-in-multiple-blanks, matching, multiple-answers, and multiple-dropdown questions, Canvas may mark every one of them answered at the start of the attempt because it saved the whole page. The guide says that log is not an accurate sequence for those question types.

New Quizzes has a different screen. On the Moderate page, Instructure documents the time a student began and completed the quiz, the events during the attempt, how long each event took, and basic event details. The page carries the same warning: do not use the log to validate academic integrity or identify cheating. It does not, in that guide, define a “stopped viewing” event, so do not assume a New Quiz recorded every time the browser lost focus. Submission timestamps on an ordinary assignment are thinner still. They say when a file arrived. They do not say who typed it, or whether a model drafted it an hour earlier.

What separate proctoring tools may record

A lockdown browser or a remote proctor is not a quiz log with a stricter name. It is software the institution chose, often launched from the quiz as an external tool. Instructure’s community reply names Respondus and Honorlock as examples of products built for integrity checks. Your campus may use one of those, both, or something else, and it may use them only on some exams.

What is recorded — a locked browser, a webcam, a room scan, a screen — is in that product’s own documentation and in the notice your course is supposed to give you. This page will not guess, and it will not describe ways around a monitor. If a proctoring session is required, the honest reading is the one in the exam instructions. A Canvas quiz log from a different assignment does not tell you what the proctor stored.

Google Classroom and Blackboard, briefly

Google Classroom’s originality reports, when a teacher turns them on, compare Google Docs, Google Slides, and Word files with web pages and books. If an administrator enables school matches, the teacher’s report can also compare against earlier student work, and Google limits that school-match feature to Education Plus and the Teaching and Learning Upgrade. The report flags passages that resemble uncited text. Students can run up to three reports on their own file before turning it in; the teacher’s report runs at turn-in. Those student help pages do not describe a ChatGPT detector. An uncited paragraph from the web can be flagged. A paragraph a model wrote, that matches nothing in the index, is a different problem those pages do not claim to solve.

SafeAssign, in Anthology’s student help, “automatically conducts the check by comparing your assignment against a database of other assignment submissions.” That page does not mention ChatGPT or an AI-writing score. Do not read the silence as a claim about every other Blackboard feature. It only means this originality check, as documented there, is an overlap check. A course that also enables a different vendor has that vendor’s report, not a new SafeAssign behavior.

What to do if an assignment is flagged

Ask which system produced the flag before you argue with it. A similarity match, an AI-writing percentage, a quiz-log row, and a proctoring incident are four different records. Only the first two are about the wording of the file. The quiz log is about focus and timing, and Instructure says not to treat it as proof of cheating. The proctor’s record is whatever that product was set to collect.

If the flag is an AI-writing percentage, read the sentences it marked and compare them with the drafts you kept. How to show the work is yours is the practical version: version history, notes, and a short account of what you did. If the awkward part is that you used a grammar tool and do not know whether that counted, what Grammarly is and is not separates a comma suggestion from a generated paragraph. Neither page can cancel a campus decision.

• Save the file you uploaded and the earlier versions, with dates.
• Read the syllabus section on outside help before the meeting, not during it.
• Ask which product scored the work, and ask to see the marked passages rather than a single percentage.
• Explain the process you used. Do not shop for a friendlier number and call it a reversal.

FAQs

Can Canvas detect ChatGPT on its own?

Not according to Instructure’s quiz-log guide or its description of how Canvas logging is designed. An AI-writing percentage on an assignment comes from a connected product. Turnitin says that product can be Turnitin, on native Canvas assignments, when the school’s license includes it.

Does “stopped viewing the quiz” mean I used ChatGPT?

No. On a classic quiz it means the quiz-taking page lost focus. The log does not record the other tab. Instructure says quiz logs should not be used to identify cheating. Some question types are also marked answered early because of autosave, so the timeline can be wrong.

Will I see the AI score in Canvas?

Turnitin says the indicator can be viewed through the LMS, including Canvas, where the institution has enabled it. That does not promise a student-facing link on your assignment. If you cannot see a report, ask the instructor rather than assuming no check ran.

Can Google Classroom detect ChatGPT?

Its originality report compares Docs, Slides, and Word files with web pages and books, and optionally with other student work at schools that have school matches. Google’s help pages describe uncited overlap. They do not describe a detector that identifies ChatGPT.

Does a quiz log prove misconduct?

Instructure says no. The classic guide and the New Quizzes Moderate page both say quiz logs should not be used to validate academic integrity or identify cheating. They can still be the start of a conversation about a timing problem.