Automatic quiz generation uses AI — typically an LLM prompted with a transcript, slide deck, or document — to draft assessment questions such as multiple-choice, true/false, short-answer, or fill-in-the-blank items. The main production workflow is: ASR produces a transcript from the video, the transcript is chunked and sent to the model with a question-generation prompt, the model returns candidate questions with distractors, and a subject-matter expert reviews and edits before the items go live. This pipeline can reduce the time to produce a first draft of a quiz from hours to minutes, which matters when a course library is large or content is updated frequently. The critical quality concern is that LLMs can generate questions that are grammatically fluent but factually wrong, ambiguous, or trivially easy, so the human-review gate is not optional. Questions generated purely from the surface text may also test recall rather than understanding; instructional designers should prompt for higher-order items explicitly. Quiz generation feeds in-video quizzes, which pause playback at the moment relevant content was covered, increasing retention. It also feeds adaptive learning systems that need a bank of items at varying difficulty to route learners effectively.
