How to Prevent Cheating on Exams: Randomization and Multiple Versions
Reduce exam cheating risks with reviewed paper versions, clear rules and careful administration. Learn what randomisation can and cannot establish.
If you are looking for ways to prevent cheating on exams, start with clear rules, appropriate assessment design and a workable administration plan. Different paper versions and randomisation can reduce some opportunities for copying. They cannot make an exam cheat-proof, prevent all collusion or prove that two papers are equally difficult.
This guide focuses on preparing and administering paper exams. Examiar creates papers and answer keys; the workflow described here is not an online proctoring, lockdown-browser, identity-verification or automatic misconduct-detection service.
Choose controls for the actual risk
| Control | May help with | Does not establish |
|---|---|---|
| Change question order | Position-by-position copying or a shared numbered answer list | That students cannot recognise or share the same underlying questions |
| Change eligible MCQ option order | Sharing an answer letter without its answer text | That the content is secret or the reordered item is still well written |
| Select different questions | Reducing overlap where the bank and selection rules permit | Equal difficulty, equal response time or protection against prior exposure |
| Clear instructions and administration | Making permitted resources, conduct and procedures explicit | That a rule or acknowledgement eliminates misconduct |
Use proportionate, institution-approved measures together. The right controls depend on the room, assessment purpose, accessibility requirements and consequences of the decision, not a universal number of paper versions.
1. Explain the rules before the assessment
State what materials, devices, collaboration and AI assistance are permitted, how students can ask for clarification and what happens if an issue is reported. Keep rules consistent with the taught work and applicable institutional or awarding-body policy.
Cornell’s academic-integrity guidance emphasises clear expectations, meaningful assessment and support for students. These are broader responsibilities than changing question order. A statement of rules is not evidence that every student will follow them.
2. Shuffle only when the question still works
Changing the order of independent questions can make a numbered answer list less useful. Keep shared passages, diagrams and dependent subquestions together. Check that instructions and cross-references still refer to the right material after reordering.
For MCQs, verify the correct answer text as well as its letter. Do not blindly reorder numerical sequences, chronological choices or options that refer to other positions. Wording such as “all of the above” needs an item-specific review; leaving it in one position does not automatically make every other permutation sound.
Review each resulting item and its key. For single-select questions, confirm one defensible best answer and plausible alternatives. See writing MCQs for the underlying design checks.
3. Review alternative questions, not just their tags
Drawing different questions may reduce overlap, but it does not mean every candidate receives unique content. A small eligible pool, repeated use and narrow selection rules can leave substantial overlap. Students may also share a solution method rather than an answer letter.
A test blueprint helps specify intended coverage and demand. It does not guarantee equal difficulty or score equivalence. Matching topic, marks and teacher-assigned difficulty labels is not enough to establish that replacement items are interchangeable.
- Compare the actual content and thinking required by the alternatives.
- Check reading load, working time, response length and access requirements.
- Verify that the marking guidance applies to each version’s questions.
- Review exposure risks and permitted reuse under the relevant policy.
- Use evidence appropriate to the stakes before treating scores from different forms as comparable.
The alternate-forms guide explains why matching a specification and establishing comparability are different tasks. A well-maintained bank supports review; it does not certify the questions inside it.
4. Release the right paper with the right key
Versioning adds responsibilities as well as options. Before printing or distributing a paper:
- Give each released form an unambiguous identifier and retain the exact paper/key pair.
- Check question order, numbering, option labels and supplied answers together.
- Reconcile attempted marks, including any internal choice.
- Review instructions, shared context, diagrams, languages and page breaks.
- Plan how papers will be distributed and how the form identifier will be recorded for marking.
- Limit access to unreleased material according to staff responsibilities.
- Use the required review process and close unresolved issues before release.
Use the moderation checklist and paper accessibility checklist to record those checks. Do not assume generating another form also completes its academic or operational review.
5. Do not substitute restrictive settings for a fair assessment
An online assessment has different risks from a supervised paper exam. A question pool, tight timer or disabled backtracking is not proof against outside help, another device or collusion. Such restrictions can also affect how students plan and demonstrate their work.
If online delivery is required, evaluate that separate system’s capabilities, accessibility, privacy and incident procedures. Do not infer those features from a paper generator. Choose timing from the actual task and required arrangements, not a goal of leaving no time to think or check work.
Reasoning or contextualised tasks may provide useful evidence of learning, but they are not inherently search-proof or AI-proof. Select tasks for the outcomes being assessed and explain the permitted support.
6. Handle concerns through the agreed process
A matching wrong answer or unusual score pattern is not, by itself, proof of cheating. Preserve relevant evidence, check for ambiguous questions or incorrect keys, and follow the institution’s reporting and review process. Do not use an automated flag or a paper-version mismatch as a substitute for investigation.
Where Examiar fits
Examiar organises questions and creates papers and answer keys. Use the pattern-design guide and documentation to prepare a small first paper, then verify the actual output and any changes you make. The software does not certify fairness, equivalent scores or exam security.
Try Examiar free with a small, non-confidential sample to evaluate the preparation workflow. Keep administration, invigilation and misconduct decisions within your approved institutional process.
Frequently asked questions
Do multiple versions stop cheating?
They may reduce particular copying opportunities, but they do not prevent all collusion, prior exposure or unauthorised assistance. Each version also needs its own review.
How many versions should a class use?
There is no universal number. Consider seating, the eligible question pool, administration and marking capacity, accessibility and the applicable policy.
Does a common blueprint make versions equally difficult?
No. It specifies intended coverage and structure, not proven difficulty or score equivalence. Review the actual tasks and use suitable evidence for consequential comparisons.