AI in the independent classroom: keeping teaching and assessment connected

The educational promise reaches the submitted work

A school may introduce AI to help a pupil explore an argument, practise a skill or receive feedback. The resulting fluency can make a lesson feel successful before anyone has established what the pupil understood. When the same pupil submits assessed work, that uncertainty becomes an institutional responsibility. The useful leadership question is how the school’s teaching approach preserves evidence of independent learning.

The distinction is particularly important now that JCQ's non-examination assessment instructions effective from 1 September 2026 sit alongside the current centre regulations and earlier AI guidance. These documents help a school turn its educational promise into assessment conditions that staff and pupils can explain. They do not amount to a new October law on AI.123

Start with the qualification and the version

The current JCQ non-examination assessment document applies to the specified GCE AS, A level and GCSE qualifications accredited in England or Wales. Its requirements are connected to the subject and assessment component. A school offering several qualification routes needs the applicable awarding-body material beside its general AI policy.1

Pearson’s May 2025 assessment update identifies the JCQ AI revision dated 30 April 2025. That earlier date should remain visible even when discussion of AI feels new. The September 2026 instructions provide a current assessment reference; a school should not relabel the older AI document as a newly issued October rule.3

Authentication begins during learning

JCQ requires centres to have arrangements for authenticating assessed work. Its AI guidance addresses acknowledgements, retained generated material and teachers’ investigation of doubts. These are connected activities: a final declaration has value when the process behind it gives the teacher a reasonable basis for accepting the work.127

That process needs to make sense beside the school's classroom offer. A pupil encouraged to explore an approved tool in a supported lesson may face different conditions in a particular assessment. Explain the change before the work begins: which assistance is allowed, what must be acknowledged and which thinking, writing or production must be the pupil's own. An ambitious approach to teaching becomes more credible when pupils can understand those boundaries.

Teacher judgement remains part of the evidence

The September non-examination assessment instructions state that teachers must not use AI as the sole or primary means of marking. Teachers must examine the whole work and determine the mark through their own professional judgement. The requirement concerns responsibility for the assessment, including when a tool offers a persuasive suggested answer.1

A governing body therefore needs a more informative explanation than whether a marking product is switched on. Subject leaders should be able to describe its permitted purpose, the work teachers still examine, how errors are identified and which decision remains theirs. A time-saving claim does not establish that feedback is accurate or that the pupil’s next learning step has improved.

Coursework connects pedagogy with secure records

JCQ’s current coursework instructions add the handling of candidates’ work to the story. They address regular backups, secure storage on centre systems and appropriate protection of electronic work. They also make clear that the instructions sit alongside awarding-body requirements. A school cannot judge an AI workflow only by the quality of text it generates.4

The digital lead and examinations officer therefore need a shared account of where drafts, declarations, feedback and supporting material are held. Keeping the final file is not enough if a platform change or disrupted account removes the surrounding record that helps explain how it was produced. Secure storage and usable records support the teacher's judgement as well as the administration of the assessment.

Current General Regulations for Approved Centres place responsibilities on heads of centre, senior leaders and examinations personnel. The 2026/27 changes also require evidence of completed cyber-security training to be held for inspection. This connects assessment integrity with staff capability and the systems used to administer it.2

Product safety and educational benefit need separate tests

DfE’s generative-AI product standards, updated in January 2026, are mainly intended for developers and suppliers. They identify intended users, purposes, testing, privacy and safety features that schools can examine when considering a product. They do not amount to a school-specific finding that a particular tool improves learning.5

The department’s education position statement likewise distinguishes opportunities from the still-developing evidence about pupil use. It places professional judgement and safeguards within adoption. For an independent school, an innovation claim should therefore identify the age group, subject, intended learning benefit and limits of the proposed use.6

A proportionate evaluation might compare pupils’ explanations, misconceptions and later independent performance before and after a defined activity. That is a proposed method, not evidence supplied by this article. It would be more informative than counting generated worksheets or assuming that a favourable response from staff establishes educational impact.

The independent-school decision sits across roles

The decision needs academic, operational and financial understanding in the same conversation. Subject leaders can explain the assessment, IT staff the service, and the bursar or governing body the continuing commitment. The head needs to see how those contributions support the school's educational proposition. These are suggested contributions to the decision, rather than a universal allocation of legal functions.

Include learning support from the outset. Assistance that removes a barrier in one activity may change the skill being assessed in another. Pupils need a clear explanation of permitted support, and staff need to check the relevant assessment and access-arrangement requirements. An appealing demonstration cannot settle a particular pupil's eligibility or establish that a tool is suitable for every task.

A school might, for example, examine whether pupils can explain and revise an argument after working with a tool, then attempt a related task independently. That is a suggested enquiry, not a tested intervention. Its value would depend on a defined learning aim, appropriate safeguards and an honest account of what pupils subsequently demonstrated.

Make the educational promise visible in the assessed work

An independent school's AI strategy becomes convincing when teaching choices, authentication, marking and secure records reinforce one another. A broad statement of ambition leaves too much unresolved if the pupil, teacher and examinations officer cannot explain how it applies to the same piece of work. The documents establish the requirements to examine; they do not show how consistently schools meet them.

The next useful evidence is the work itself and the judgement around it: a defined use, the teacher's role, what pupils could explain afterwards and what proved unsuitable. Keep the rejected uses and difficulties in that account. The educational case grows stronger when the school can show what pupils learned, with assessment records that support the conclusion. No original classroom observation or school-level outcome study was undertaken for this feature.