AI in Vocational Assessment: Redefining Formative Assessment for 2026

This matters most in UK vocational training, where assessment must still produce valid, reliable and genuine evidence of a learner's knowledge, skills and understanding. Used well, AI is not only a threat to that aim. It is a chance to build assessment that is more meaningful, more transparent and closer to real practice.

Generative AI can now produce written explanations, lesson plans, summaries and presentations. The Department for Education (DfE) recognises that it can support tasks such as resource creation and tailored feedback, while also flagging risks such as inaccuracies and the misuse of personal data. ^[3] Assessors therefore need to understand both the benefits and the limits of these tools, and design assessments that encourage responsible use.

The difficulty is that a polished submission can reveal very little about the learning behind it. A well-written document may say nothing about:

• how the learner developed their ideas
• the research they actually undertook
• their real comprehension of the subject
• how they responded to feedback
• how they would apply the knowledge in practice
• how they evaluated different sources
• the decisions they made along the way

So the critical inquiry goes beyond 'Did the learner use AI?' It becomes:

Assessors. Your professional judgement matters more, not less. Confirming that evidence is the learner's own now means looking at how it was produced, not only at how it reads.

Internal quality assurers. Sampling should cover the evidence trail (drafts, declarations, observation notes) as well as the final product. Standardisation meetings are a good place to agree how AI declarations are reviewed and what triggers a follow-up conversation with the learner.

Learners. Clear rules at the start of each assignment protect learners too. When they know what is allowed, what to declare and what to keep, they are far less likely to drift into malpractice by accident.

Formative assessment takes place during learning and gives both learners and educators insight into progress. Questioning, quizzes, discussions, draft work, peer assessment, feedback, observation, reflective activities, practical demonstrations, presentations and self-assessment all reveal thinking that a finished document cannot.

Generative AI doesn't make these methods obsolete. It increases their value, because educators can focus more on process, application, explanation, reflection and authentic performance. Take a familiar task:

The result is a far more complete picture of learning than a single submitted document.

AI invites us to rethink the role of the final written submission. Well-crafted documents are valuable, but they may not capture the full range of knowledge and skills a qualification demands. Give learners more chances to show:

• Critical thinking: why they chose an approach, and where they challenged an AI output.
• Evaluation skills: how they compared and judged sources.
• Practical application: using knowledge in a realistic scenario.
• Reflective insight: what they learned and what they would change.
• Clear communication: explaining their work aloud, not just on paper.
• Practical competence: doing the job, observed in real time.

These methods are especially relevant in vocational education, where practical skills often need to be demonstrated beyond written responses.

With AI comes an amplified need for authenticity. For regulated qualifications, learners must still meet the criteria in the qualification specification and follow assessment regulations. JCQ's guidance is clear that centres should accept only work that belongs to the learner:

Misusing AI to the point that submitted work is no longer the learner's own can amount to malpractice, and where AI use is permitted it should be properly acknowledged. Which rulebook applies depends on the qualification:

If concerns arise about a learner’s submission, our guide explains what to do when an assignment is flagged as AI-generated and how to approach the situation fairly.

A practical AI declaration checklist

Requirements vary by awarding organisation, so confirm the detail in your own policy. As a baseline, ask learners to:

1. State whether AI was used
2. Name the AI tool or platform
3. Save the original prompts and unedited outputs
4. Explain how AI was used and which parts it influenced
5. Critically check AI content before including it

Also set out accepted and prohibited uses at the start of each assignment, and use more than one route to confirm authenticity rather than relying on how polished a submission looks.

1. Being vague. 'Use AI responsibly' isn't a rule. Say what is permitted, what is prohibited and what must be declared.
2. Relying on detection alone. Detection tools can't replace evidence of process and professional discussion. For more detail, read our guide on whether AI detection tools can reliably identify ChatGPT in academic writing .
3. Ignoring the awarding organisation. Its policy is the one that governs your qualification.
4. Forgetting staff training. A policy only works if assessors and IQAs apply it consistently.
5. Treating AI as only a risk. Teaching learners to fact-check and critique AI is part of preparing them for work.

AI isn't only something learners use. Jisc's National Centre for AI in Tertiary Education has been running a year-long AI in assessment project from September 2025, trialling tools that support marking and feedback in further and higher education. [4] The promise is faster, more consistent feedback and a lighter marking load. The caution is just as important: accuracy and consistency of AI-generated marking are still open questions.

If you use AI to support your own assessment workload, keep these guardrails in place:

• The assessor, not the tool, makes the final judgement.
• Sample and check AI-generated feedback before it reaches learners.
• Never paste identifiable learner work into an unapproved tool.
• Tell learners how AI is, and isn't, used in marking.

Generative AI isn't the end of meaningful assessment. It's a prompt to make assessment better. When we focus on process, authenticity, reflection and professional dialogue, a polished final document stops being the only proof of learning. Learners still need to demonstrate their knowledge, skills and understanding, and assessors still need to be confident the evidence is genuinely theirs.

The path forward is not to ban AI or chase detection tools. It is to set clear expectations, require transparency, protect learner data and design tasks that show real competence. Done well, AI becomes a tool for learning rather than a shortcut around it, and human judgement stays at the centre of assessment. To build those skills, explore our Level 3 CAVA assessor qualification or the Level 3 Award in AI in Education.