Can Canvas Detect ChatGPT for Multiple Choice? Quiz Logs, Proctoring, and What Actually Gets Tracked

Can Canvas Detect ChatGPT for Multiple Choice Questions?

Canvas is a learning management system built by Instructure for managing coursework, grades, and communications. It does not include any AI detection capability anywhere in its native platform. The AI detection experience that students sometimes encounter inside Canvas — the percentage score that appears in SpeedGrader alongside a Turnitin report — applies only to written assignments routed through a Turnitin LTI integration. Multiple-choice quizzes sit in an entirely separate part of the Canvas workflow and are never touched by that pipeline. When a student selects an answer in a Canvas quiz, they are clicking a radio button or checkbox that records a chosen option from a predefined set. There is no student-generated prose, no linguistic content for a statistical model to evaluate, and no mechanism that would route a selected quiz answer to a text-based detection service. Can canvas detect chatgpt for multiple choice in the sense that Turnitin's AI Writing Indicator detects AI-generated essays? No — and not because the technology has not been tried, but because the detection model does not apply. AI text detectors analyze perplexity, which measures how predictable each word choice is given its surrounding context, and burstiness, which measures variation in sentence length and rhythm patterns across a document. Both signals require a body of continuous written text to produce any meaningful probability estimate. A selected answer — A, B, C, or D — contains none of those signals. This distinction is foundational: the AI detection workflow for written assignments is structurally irrelevant to the multiple-choice quiz context, regardless of which institution or tools are involved.

What Does the Canvas Quiz Log Actually Record?

While Canvas cannot apply AI detection to multiple-choice answers, it does automatically generate a quiz log for every student who takes a Canvas quiz. The quiz log is a timestamped record of student activity during the exam session, and instructors can access it from the quiz submission detail view inside their course. The log records when the student opened the quiz, how much time elapsed on each question before they selected or changed an answer, and when the final answers were submitted. It also captures page-focus events: if a student's browser window loses focus during the quiz — because the student switched to another tab, opened a new application, or clicked outside the quiz window — Canvas records this as a page leave event with a timestamp. These page leave events are the closest thing Canvas has to a behavioral flag for quiz-taking. They do not record what the student viewed while outside the quiz window, what applications were running, or what text was entered into other programs. Canvas records behavioral timing data within its own interface, nothing beyond that. Answer revision patterns are also visible in the quiz log: instructors can see how many times a student changed their selection on a specific question, and how long elapsed between the first selection and the final one. For some question types, Canvas logs whether a student returned to a question after initially moving past it. Instructors who use quiz logs in academic integrity reviews typically look for combinations of signals — multiple page leave events, unusually fast answer times on difficult questions, or answer revisions that follow closely after a page leave — rather than any single data point. A pattern of leaving and returning before changing an answer carries more interpretive weight than either signal alone.

"The Canvas quiz log gives instructors a question-level timeline, not a screen recording. It tells you when and whether a student left the quiz window, not what they looked at while they were away." — Canvas LMS administrator documentation

Does LockDown Browser Catch Students Using ChatGPT?

LockDown Browser is a customized application developed by Respondus that instructors can require for Canvas quizzes. When a student opens a Canvas quiz that has LockDown Browser enabled, they must launch the exam inside the LockDown Browser application rather than in their standard web browser. The application restricts access to other browser tabs, desktop applications, and clipboard paste functionality during the exam session. A student cannot open a regular browser window to visit ChatGPT while taking a quiz inside LockDown Browser on that same machine. Whether LockDown Browser fully prevents ChatGPT access during a multiple-choice quiz is a narrower question. A student with a second device — a phone, tablet, or secondary laptop — can access ChatGPT on that device while taking the exam in LockDown Browser on the proctored computer. LockDown Browser has no visibility into other devices in the student's environment. Respondus Monitor, the companion webcam proctoring add-on, records video and audio throughout the exam session and applies automated behavioral analysis to the footage. Monitor flags segments where a student looks away from the screen for extended periods, where another person appears in the camera frame, or where significant ambient noise events occur. These behavioral flags are reviewed by the instructor or a proctoring staff member — Monitor does not automatically determine that a student used ChatGPT or any other outside resource. The flagging is behavioral, not content-based. An instructor reviewing Monitor footage would be making a judgment about whether the student's gaze and physical behavior suggest consultation of an outside source, not whether an AI-generated answer was submitted.

1. Check the Canvas assignment instructions to confirm whether LockDown Browser is required before the quiz date
2. Download and install the Respondus LockDown Browser application from your institution's student software portal
3. Test the application on a practice exam or sample quiz before the actual exam date to catch any compatibility issues
4. Close all non-essential applications before launching the quiz in LockDown Browser
5. If Respondus Monitor webcam recording is also required, complete the environment check and webcam setup before the exam timer starts

Which Canvas Proctoring Integrations Monitor Multiple-Choice Exams?

Several third-party proctoring platforms integrate with Canvas through LTI connections and can be configured to run during multiple-choice quizzes. Proctorio is among the most widely deployed Canvas proctoring tools at U.S. universities, operating as a browser extension that activates for exams configured with Proctorio in the Canvas assignment settings. The extension records the student's screen, webcam, and microphone during the exam, and generates automated suspicion flags based on detected behaviors such as extended gaze away from the screen, use of application-switching keyboard shortcuts, or additional voices in the audio. These automated flags produce a review report — they do not automatically result in an academic integrity finding. Instructors or institutional proctoring staff review flagged segments before any action is taken. Honorlock integrates with Canvas as an LTI app and combines browser extension monitoring with the option for a live human proctor to join the session if automated monitoring detects unusual activity. Honorlock has marketed secondary device detection functionality intended to identify when a nearby smartphone or tablet is actively browsing the internet, though the practical reliability and institutional use of this feature vary considerably. ExamSoft is a separately downloaded application used primarily in professional licensing and graduate credentialing contexts rather than standard undergraduate coursework; it sequesters exam content in an offline format and uploads behavioral recordings after the session. These integrations differ meaningfully in their technical capabilities and in what students can expect in terms of visibility and notification. Reading the technical documentation and any student rights disclosures your institution provides for the specific proctoring tool in use is the most reliable way to understand what that particular system monitors.

Do Timing Patterns in Quiz Logs Reveal AI Use?

Instructors who review Canvas quiz logs in connection with academic integrity concerns typically look for timing patterns that fall outside the normal range for the student cohort on that specific exam. For multiple-choice questions, answer times that are extremely short on questions that require careful reading or multi-step reasoning attract attention — particularly when those fast answers coincide with recorded page leave events. If the quiz log shows a sequence of leaving the quiz page, returning, and changing an answer on multiple questions within the same session, that pattern is more interpretively significant than either timing or page leave events considered separately. For the question of whether can canvas detect chatgpt for multiple choice through behavioral monitoring, the relevant sequence is: a student leaves the quiz window, presumably to consult an outside source, returns with an answer, and repeats this pattern across questions. Canvas records exact timestamps for when each answer was changed and when the quiz window lost and regained focus, which makes this sequence visible in the log data. Timing-based suspicion is inherently probabilistic. Students complete questions at different speeds for reasons that have nothing to do with outside assistance — test anxiety, re-reading a confusing question stem, reconsidering a calculation, or brief technical interruptions like a slow connection or a system notification. Institutions with structured academic integrity review processes generally require instructors to document timing evidence alongside other indicators — the student's course performance trend, comparison with results on previous quizzes, and the difficulty distribution of the specific flagged questions — before opening a formal review. Timing data from a quiz log is treated as one signal among several, not a self-sufficient finding.

"Quiz log timing data is one data point among many. A student who finishes quickly is not necessarily using outside resources — some students simply know the material well. Patterns across multiple questions, combined with other signals, are what justify a closer look." — Academic integrity administrator, 2025

Why Do AI Text Detectors Not Apply to Multiple-Choice Quiz Answers?

AI text detectors like Turnitin's AI Writing Indicator, GPTZero, Copyleaks, and comparable tools were developed specifically to identify AI-generated written prose. Their methodology depends on two primary statistical signals derived from continuous text: perplexity, which measures how predictable each word choice is given the words surrounding it, and burstiness, which measures variation in sentence length and syntactic complexity patterns across a document. Both signals require a body of continuous text — at minimum several complete sentences — to generate any statistically meaningful probability estimate. Turnitin itself discloses that submissions under 300 words produce unreliable AI Writing Indicator results, which gives a sense of the sample size requirement even for the written-text case. Multiple-choice quiz answers produce none of this analyzable content. When a student selects option B for question 12, the recorded response is a single letter or a short answer-label string. There are no word sequences to evaluate for predictability, no sentence structures to analyze for length variation, no perplexity score to calculate. The core question of whether can canvas detect chatgpt for multiple choice by text analysis is therefore moot: the detection model has nothing to work with. ChatGPT's role in a multiple-choice exam context, when a student uses it, is as an information source consulted outside the exam interface — not as a text generator producing output that can be statistically fingerprinted. What Canvas quiz logs and proctoring systems attempt to monitor is the behavioral evidence of outside consultation, not the linguistic content of the selected answers themselves. These are fundamentally different monitoring approaches aimed at different evidence types.

What Should Students Know Before Taking a Proctored Canvas Quiz?

Before taking a Canvas multiple-choice quiz with any proctoring or monitoring configuration, reading the assignment instructions carefully is the most useful preparation. Instructors are typically required to disclose which monitoring tools are active, and many Canvas institutions require explicit language about LockDown Browser, Respondus Monitor, or third-party proctoring tools in the assignment description or course syllabus. If the instructions mention required software, downloading and testing it before the exam date prevents technical problems during the actual exam window — LockDown Browser in particular requires installation and sometimes a system compatibility check that can surface problems early when there is time to address them. Proctoring integrations like Proctorio and Honorlock typically run a pre-exam setup sequence that includes a webcam environment scan and application check. Completing a setup check in advance, during a non-exam session, removes friction on exam day. From an academic integrity standpoint, the practical resolution to the question of what canvas records during a multiple-choice quiz is this: Canvas quiz logs capture timing and page-focus behavior; proctoring tools, when enabled, capture webcam and screen activity; secondary devices are outside the technical reach of most monitoring systems but remain within the scope of most institutional academic integrity policies regardless of detection capability. Whether a specific act of consulting an outside source would be detected depends on which tools are enabled and what behavioral evidence was generated. Whether using an outside resource falls within the rules of a specific exam depends on your institution's AI policy and your instructor's explicit exam rules — and those rules apply even in the absence of technical detection.

1. Read the Canvas quiz instructions for explicit language about required monitoring software before the exam date
2. Check the course syllabus for AI policy language that applies to quizzes, not just written assignments
3. Download and test any required proctoring application — LockDown Browser, Proctorio extension, or Honorlock — at least 24 hours before the exam
4. Complete the pre-exam setup check or environment scan for webcam-based proctoring tools before the actual exam session
5. Close all non-essential browser tabs and desktop applications before starting the quiz
6. Ask your instructor in writing if the assignment instructions do not specify which monitoring tools are active