FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Apex Insurance Brokers
Speak to a brokerGet a quote →

Aipedia: Training providers and educators

AI-generated exam questions and assessments

Providers can use AI to draft exam questions and case studies, but must check every item for accuracy, validity and bias. A flawed assessment affects every candidate who sits it and can produce grouped claims.

In short

AI can draft question banks, case studies and mark schemes quickly. Ofqual sees potential but flags accuracy, bias and appropriateness risks, is researching AI-generated items and will issue further guidance. Awarding organisations stay responsible for valid, reliable and fair outcomes. Possible failures include wrong keyed answers, questions that test the wrong thing, culturally biased scenarios and gaps in coverage. Because one flawed item affects many candidates, claims can arrive together and aggregation matters. Expert sign-off, piloting and bias review are the main controls.

How providers use AI to write assessments

Writing good assessment items takes time and skill. Providers may use AI to draft multiple choice questions and distractors, short answer prompts, case studies, scenario-based tasks, mark schemes and question banks for online tests. Corporate academies may use AI to generate end-of-module quizzes at scale, and Ofqual is researching AI in assessment design.

The appeal is obvious: a tool can produce fifty plausible questions in the time a writer drafts five. But a plausible question is not the same as a good one. Assessment items have to measure the right thing, have one defensible answer, be fair to every group of candidates and stay secure. Those are judgements AI does not reliably make, and errors may be invisible until candidates start appealing.

What Ofqual says

Ofqual's approach to regulating the use of AI (republished 16 July 2026) sees AI in assessment design as a source of innovation, but flags risks relating to accuracy, bias and the appropriateness of AI-generated content. It says it is researching AI-generated assessment items and will issue further guidance on good practice. Awarding organisations remain responsible for valid, reliable and fair outcomes, with expert human oversight.

For other providers, the DfE principle in its generative AI in education guidance that staff check AI output and remain responsible for it applies just as well to a quiz as to a handout.

Validity, accuracy and bias risks

What can go wrong

Illustrative scenario: a provider of a licence-to-practise course uses AI to expand its online question bank. Several new items have incorrect keyed answers. Pass rates drop, candidates appeal, and an employer that funded a cohort complains that staff could not start work on schedule. The provider has to regrade results and faces a claim for the employer's costs.

Because a flawed question affects every candidate who sees it, the claims may come in groups. How your policy aggregates related claims then matters; see aggregation of claims in PI.

How PI usually responds

A claim alleging that a provider negligently designed an assessment, causing financial loss, may fall within a trainers' PI policy, subject to its terms. Some policies treat awarding or certification work separately or exclude it; see training consultancy and certification insurance. Costs of re-running or regrading an assessment yourself are not usually covered unless they reduce a covered claim.

Practical controls

How Apex helps

Apex arranges PI for training providers and assessment bodies, and a broker can check how your policy treats assessment design and certification. See the training providers Aipedia hub, AI marking and feedback, AI-generated course materials and the Ofqual, Ofsted and DfE rules explained.

Sources

Frequently asked

Can we use AI to write exam questions for a regulated qualification?

Ofqual does not ban it. It sees potential but flags accuracy, bias and appropriateness risks, is researching AI-generated items and plans further guidance. The awarding organisation remains responsible for valid, reliable and fair assessment, so expert review of every item is essential.

What if a flawed AI-written question affected results?

Correct the item, review affected results promptly and tell candidates what you are doing. Keep records. If candidates or funders allege loss, notify your insurer as a circumstance where the policy requires, as several related claims may follow.

Does PI cover the cost of rerunning an exam?

Usually not as a standalone cost, because PI responds to claims by third parties. Some policies may contribute to costs that reduce a covered claim, known as mitigation costs, depending on the wording. Ask your broker how your policy treats this.

Should we tell candidates that questions were written with AI?

There is no general rule requiring it, but awarding bodies, funders or corporate clients may ask how assessments are produced. Being able to show that every item was reviewed and approved by a subject expert matters more than the drafting tool. Keep that record for each item.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.