This is the operational companion to the case for AI exam import: the exact flow from document to published quiz, the settings that matter, and the fixes for the messy formats that defeat simpler tools. Total hands-on time for a typical 60-question paper: under ten minutes, most of it review.
The five-step flow #
- Upload: drag the PDF, Word file or scanned images into Exam Import
- Parse: the AI reads structure — questions, options, answers — and shows a preview
- Review: fix flagged items; low-confidence questions are marked for you
- Configure: set question types, points, shuffle rules and feedback visibility
- Publish: the quiz lands in your course — or in your question bank for later
When the PDF fights back #
Real exam papers come in hostile formats: two-column layouts, answer keys on the last page, sections with different numbering, headers that repeat every page. The importer handles most of this automatically — it detects section boundaries and reads keys wherever they sit. When something does misparse, prefer the surgical fixes: merge a question split across pages, reassign an option that attached to the wrong item, and mark true/false statements the layout hid inside prose. Two minutes of surgery beats re-typing every time.
Settings that matter #
Before publishing, three decisions shape the student experience: feedback timing (immediate for practice, post-submission for exams), shuffle (on for assessment integrity, off when question order follows a narrative), and pool behavior — whether the quiz shows all imported questions or draws a random subset per attempt. That last one converts a static imported paper into an unlimited-practice generator: 60 imported questions drawing 15 per attempt gives every student a different, repeatable drill.