Today's AI models write MCQs riddled with test-wise cues—the longest option, the oddly specific one—that let unprepared students guess their way to a pass. TiobExam is engineered to strip those cues, so a passing score means real knowledge.
Built by a medical educator with 15+ years of exam-writing experience
"I spent years hand-writing questions to keep them fair. I built the tool that does it at scale—and proves it works."
155 Level-300 medical students were examined on material they had not yet been taught. With no real knowledge, their scores should sit near the 25% guessing line. On AI-generated questions, they didn't.
Every student sat all the question sets. The same student who scored 86% on a leading AI model's questions scored 38% on ours—because AI-generated questions leak the answer through test-wise cues, and ours don't. The inflation isn't learning. It's the longest option, the oddly specific phrase. And it isn't one chatbot—general-purpose AI models share the same blind spot.
Diagnostic study of 155 Level-300 medical students, examined on material not yet taught. Full analysis available on request.
Anyone can generate a question bank in minutes now. The problem is what those questions quietly do to your results.
You prompt an AI for 30 MCQs. They come back polished, formatted, and ready to use.
The correct option is the longest. Or the most specific. Or the only one without an absolute term. The structure points straight at the key.
They never studied the material—but they read the cues, not the content, and score 80–90%.
The result your accreditation body trusts—and, in medicine, that patients depend on—no longer reflects what students actually know.
TiobExam generates the questions and guards their integrity—then runs your whole exam operation around them.
Questions are engineered so the structure never points to the answer. Students can't pass by spotting patterns.
In our study, students who hadn't learned the material couldn't beat our questions—unlike AI-generated ones.
Years of assessment research, distilled into a quality engine that guards every question's integrity.
Bloom's-aligned questions from your lecture materials—higher-order thinking, in minutes.
Assign quotas, auto-remind lecturers, and track every submission across a multi-lecturer course.
One click to Moodle/GIFT, PDF, or Word. No reformatting ever again.
Anyone can generate questions. TiobExam goes further—screening every item for the cues that inflate scores, upgrading the questions you already have, and assembling them into balanced, defensible exams.
Before an item is ever shown to a student, it passes through our quality engine—built to catch the answer-revealing patterns that let unprepared students guess their way to a pass.
Our questions can't be cracked by spotting the longest or oddest option. The structure never gives the answer away—so a score reflects what a student actually knows.
We validated it on students who never studied the material—they couldn't beat our questions. And every general-purpose model we evaluated produced the same gameable cues; ours scored highest for quality.
Every item is checked against the principles of sound assessment design and flagged if it falls short. The methodology is the result of dedicated research—and it's ours.
Already have an exam bank? Upload it and get an instant quality report—each question scored, its weaknesses flagged, and clear fixes suggested. Your existing questions, held to the same standard.
Turn old or rough questions into polished, standards-aligned items. Choose how much changes—a light polish or a deep rework—and keep full control of every result.
Define a Table of Specifications once—topics, cognitive levels, and difficulty—then assemble balanced exams from your curated Question Vault, with live coverage tracking and auto-filled gaps.
Choose your cognitive framework—Bloom's Taxonomy or Webb's Depth of Knowledge—and set the level distribution up front. Bloom's measures the type of thinking; DOK measures its depth. Either way, you target real rigor, not just recall.
Assign question quotas, auto-remind lecturers at 7/3/1 days, track every submission, route external examiner review, and approve—one dashboard for a multi-lecturer course.
Export approved questions to Moodle (GIFT), Aiken, PDF, or Word. No reformatting, no copy-paste errors. One click and you're done.
A streamlined workflow that respects academic expertise—with a quality gate that does the part a busy lecturer can't do by eye.
Upload lecture slides, notes, or PDFs—or start from learning objectives alone. Advanced OCR extracts content from even scanned documents and image-heavy presentations.
AI drafts Bloom's-aligned questions from your content—then the real work begins. Every item passes through our quality engine, which removes the hidden cues that let students guess and flags anything that doesn't measure real understanding. Weak items are rewritten automatically.
You stay in full control. Review every question with the quality report attached, edit as needed, approve or regenerate. Export directly to Moodle, PDF, or Word when ready.
Includes AI generation, our quality screening, and your review. Quality you couldn't do by hand—at a fraction of the time.
In a 30-minute demo we'll generate questions from your actual materials—and show you the quality engine flag and rewrite a cued question live.
Generate → Screen → Export