AI quizzes your faculty can defend.

Faculty are using ChatGPT to generate quizzes. You know it. The committee knows it. Nobody's saying it out loud because nobody wants the conversation that comes next when a student appeals a grade and the instructor can't explain where a question came from.

You have three options:

1. Ban it — futile, faculty will use it anyway, and you lose visibility.
2. Wait for the LMS vendor to ship something — Canvas / Blackboard roadmaps are 18 months out, and what they ship will be black-box.
3. Adopt a tool that bakes traceability in from day one.

QuizCraft is the third option.

Every question QuizCraft generates is anchored to a specific paragraph in the source material the instructor uploaded. Not a vague topic match — the literal sentence the answer comes from, included with the question and exported alongside it.

• The instructor reviews citations before publishing — catches the occasional bad question in seconds, not after a student complaint.
• The exported QTI (Canvas / Blackboard / Moodle / Brightspace) includes citation metadata — so a year from now, an integrity review can trace any question back to its source.
• No source material is ever used to train AI — written into the contract, not just the Privacy page.

Eval methodology and full numbers below. We re-run this eval before every prompt change.

• Not vendor-locked — instructor library stays exportable forever, even after the contract ends. If the pilot ends, we hand back every file and delete the rest.
• Not a proctoring tool — no surveillance, no student data.
• Not a black-box AI — every output is reviewable and editable by the instructor before students see it.
• Not an LMS — works alongside Canvas / Moodle / Brightspace / Blackboard.

All tiers: invoiced annually, NET-30 terms, no individual credit cards required.

The window where “early adopter of responsible AI assessment” is a positioning win for a CTL is shrinking. By Fall 2026, every LMS vendor will be marketing AI features and the conversation shifts to “which vendor's AI is best”, not “is your CTL leading on this?”

A 25-seat pilot is a $1,475 budget line and a 4-week commitment to findings. It doesn't require IT approval (Google SSO only, no SAML needed for pilot), doesn't trigger procurement, and gives your CTL concrete data for the harder conversations coming next year.

We don't want you to take the 100% number on faith. Here is the exact procedure:

1. Three hand-curated teaching passages (intro-college level) across macroeconomics, US history, and software engineering. Each ~300-400 words, dense enough to mirror real reading assignments.
2. 5 questions generated per passage (MCQ + True/False + Fill-in-the-blank), totalling n=15 questions.
3. Programmatic check #1 — quote retrievability.For each question, we look for the cited source quote in the original passage. Exact substring match, or fallback to >80% word overlap (for paraphrased citations).
4. Programmatic check #2 — distractor balance.For MCQs, flags “correct answer is conspicuously the longest” (a notorious AI-quiz tell) and lazy distractors like “all of the above.”
5. Claude judge — answer support.An independent Claude call evaluates whether the cited quote actually semantically justifies the answer. Strict: a teacher should be able to point at the quote and say “see, the answer is here.”

Results above are from the most recent methodology run. We're happy to share the full eval — script and all — by email, including which specific questions fell in the 13% “partial” bucket. Transparency on what we're NOT claiming matters more than the headline.

The fastest two paths:

• Try it yourself in 60 seconds, no signup: usequizcraft.com/try — runs against a fixed sample reading; see citations populate inline.
• Send us a PDF from one of your courses — we generate a quiz from your actual material and walk you through the citation flow in a 15-minute call. Email yaoyanweb@gmail.com.

QuizCraft is built by Connor Yao, a solo engineer who's spent six months making AI quiz generation actually defensible — not just fast. Direct line: yaoyanweb@gmail.com. Every CTL inquiry gets a personal reply within 24 hours.