AI Question Generation

Generate IB-style questions on demand — grounded in the official syllabus where available, calibrated to your chosen difficulty and command term, and saved to your personal question bank for assignments and practice.

What it does

AI Question Generation drafts new IB-style questions on a topic of your choice. You give it a subject, a subtopic, a question type, a command term, a difficulty and a mark allocation; the model produces a question stem (plus mark scheme where applicable). You review, edit if needed, then save to your personal question bank for use in assignments or for your students' practice queues.

Mark-scheme aware

For exam-style questions the generator drafts both the question and a multi-criterion mark scheme aligned to IB command terms.

Difficulty-calibrated

Difficulty is requested explicitly (Foundation, Intermediate, Proficient, Advanced) and reflected in the stem complexity and expected response depth.

Subtopic-targeted

You pick the subtopic, not just the subject — so a Biology cell biology generation won't drift into ecology.

Owned by you

Generated questions are tagged with your email as the author and only appear in your bank. Students see them only when you include them in an assignment.

Where to find it

AI Question Generation is in the Content section of the teacher sidebar, alongside Question Bank, Assignments, and Mark Answers. The Question Bank page also has a "Generate new" shortcut that drops you straight into the generator with the bank's current subject and subtopic pre-filled.

Teachers only.

This feature is gated to accounts with the Teacher role at a school. Students can't reach it. Every teacher gets the same monthly quota — see Monthly quotas.

The generation workflow

Set the parameters

Pick a subject, a subtopic, the question type (Multiple Choice, Fill-in-the-Blank, or Exam Style), a command term (State, Describe, Explain, Analyse, Evaluate, Compare, etc.), a difficulty, and the marks the question should be worth. For Exam Style, the marks drive the expected length and the mark-scheme structure.

Optional context prompt

Add a one-line steer if you want the question targeted at a specific scenario ("about renewable energy in a developing-country context") or to avoid certain content ("not about ATP synthesis — we covered that last lesson"). The generator will incorporate this into the syllabus-grounded prompt.

Generate & review

Click Generate. The AI returns the question stem (and mark scheme for exam-style) in a few seconds. Review it on the right-hand pane. The usage widget on your dashboard increments by one as soon as the call succeeds.

Edit if needed

Both the question text and the mark scheme are fully editable. Make any tweaks — rewording, removing a misleading distractor in an MCQ, adjusting the mark scheme's wording — before saving.

Save or regenerate

Save commits the question to your bank. Regenerate re-runs with the same parameters (uses another quota credit). Discard drops it without saving (also counts against quota, since the generation already ran).

Syllabus grounding

For subjects we've onboarded, the generator runs in syllabus-grounded mode: it injects per-subtopic learning outcomes, key concepts, SL / HL level markers, common misconceptions, example phrasings, and linked topics into the prompt before drafting. The effect is that command-term-appropriate, IB-style language is far more consistent and topic drift is rare.

For subjects without grounding yet, the generator still runs — it falls back to general IB style cues from the subject name and command term. Grounded subjects are clearly marked in the subject dropdown with a small badge.

Coverage is expanding.

Computer Science was the first subject to ship with full syllabus grounding. Other subjects are rolling out as their syllabus content is curated. Until they're all onboarded, expect noticeably higher quality on grounded subjects.

Monthly quotas

Every teacher gets 25 AI generations and 50 AI marks per month. The quota resets on the 1st of each month and is per teacher (not pooled across the school). These caps bound shared AI capacity; they aren't a paid-tier thing.

The usage widget

Your TeacherDashboard has a usage widget under the welcome banner showing X / N AI generations for the current month and the reset date ("Resets [date]"). The widget bar turns amber at ≥80% utilisation.

What happens when you hit the cap

The generator returns a structured soft-block message: "You've used all 25 AI generations this month. Resets [date]." The page surfaces this as a banner; the question itself isn't drafted, and the quota isn't incremented. Existing draft questions, assignments, and your question bank are unaffected — only further AI calls are gated.

What counts against your quota

Every successful generation (whether you save the question, regenerate, or discard it).
Generating a new mark scheme for an existing question.
Regenerations — each one is a fresh call.

What doesn't count: editing a saved question's text manually, deleting questions, using questions in assignments, marking student work.

Using generated questions

Saved questions land in your personal question bank. From there:

Assignments. In the assignment wizard's Step 2 (Type & Questions), pick Manual mode and toggle "Showing your own questions" — the filter scopes to questions you authored.
Student practice. Tag the question to make it discoverable in the student-side practice filter, or leave it un-tagged to keep it exclusive to your assignments.
Edit or retire. The bank supports inline editing and soft-deletion. Questions used in past assignments stay attached to those assignments even if you retire them later.

Quality tips

Be specific in the subtopic. "Cell respiration" is better than "Biology". Subtopic specificity is the single biggest lever on quality.
Pick the right command term for the marks. A 1-mark State question and an 8-mark Evaluate question are structurally different; matching them properly avoids mark schemes that feel padded or rushed.
Use the context prompt for variety. Generating five "Explain photosynthesis" questions with no extra context will produce variants that feel similar. Add a one-line steer ("set in the context of crop yields") and the generator diversifies.
Always review before assigning. The generator is good but not perfect — treat its output as a first draft. A 30-second sanity check before saving is well spent.
Mark schemes are starting points. Tighten the wording to match your school's marking conventions before saving. The AI marker on student answers will follow the saved mark scheme verbatim.

Troubleshooting

"Quota exceeded" banner
You've used all of this month's allocation. Wait until the reset date shown in the widget, or talk to your school admin about upgrading. Existing questions still work.
The question drifted off-topic
Regenerate with a more specific subtopic, or add a context-prompt steer. If a particular subtopic consistently misfires, let us know — it usually means the syllabus grounding for that area needs a tweak.
Mark scheme too generic
Lower-stakes question types (MCQ, fill-in-the-blank) auto-generate compact schemes; for richer schemes, switch to Exam Style and pick a higher mark count. You can also edit the mark scheme in place before saving.
Empty error banner with no message
That was a known bug on a previous version; if you still see it, refresh and retry, and contact support@ibmarker.com if it persists.