Using AI for IB / IGCSE / A-Level exam academic writing

Artificial intelligence has created a rather peculiar problem for English students. Used well, it can provide instant feedback, generate practice material and help students interrogate their own ideas. Used badly, it can produce an impressive essay that teaches the student almost nothing.

The important question, therefore, is not simply whether students should use AI. Many already do. The more useful question is: how can students use AI and become less dependent on it over time?

The Problem with the Instant Answer

Students commonly use tools such as ChatGPT to summarise texts, explain quotations, generate essay plans, improve vocabulary, correct grammar or produce entire paragraphs.

Some of these uses are valuable. Others merely disguise weaknesses.

Suppose a GCSE student is given a language-analysis question. An AI tool can produce a polished response in seconds, complete with terminology, quotations and sophisticated interpretations. The homework may improve dramatically.

But has the student improved?

The distinction matters because the examination eventually removes the support. A student who repeatedly outsources the difficult part - interpreting evidence, developing an argument, and choosing appropriate language - may become very good at producing impressive homework while remaining poorly prepared for the examination.

The aim should therefore be to use AI as a scaffold rather than a substitute.

GCSE/IGCSE: Make the Student Think First

Consider a genuine AQA GCSE English Language question asking students how a writer uses language to describe an early morning. AQA's examiner feedback shows that stronger responses explore layers of meaning, whereas weaker answers sometimes drift into general comments rather than analysing the text itself.

A student could simply paste the extract and question into AI and request a model answer.

A better approach reverses the process.

First, the student spends ten minutes identifying three interesting language choices and explaining their effects. Only then do they ask AI to analyse the same passage.

Now there is something worth discussing:

What did the AI notice that you missed? Which of its interpretations are convincing? Which are generic? Can you find better textual evidence?

The student remains responsible for judgement. AI has become something to argue with rather than something to copy.

A tutor can then ask the student to rewrite one paragraph before tackling another extract without AI.

A-Level: Use AI to Challenge an Argument

At A-Level, the danger becomes subtler because students can ask AI to generate increasingly sophisticated literary arguments.

Take an actual AQA question on Othello, which asks students to explore how and why Shakespeare presents characters' conflicting senses of duty. The mark scheme permits numerous possible approaches: marital and family duty, personal and social obligations, honour, violence and the conflicting responsibilities of characters such as Othello and Cassio.

Instead of asking AI, "Write an essay answering this question," the student might formulate a thesis first:

Othello's tragedy partly develops because his competing ideas of marital, military and masculine duty become impossible to reconcile.

AI can now be asked to attack that argument.

What evidence contradicts it? What alternative interpretation could be made? Which scenes complicate the claim?

This is a considerably more demanding use of the technology. The student has produced the argument; AI is being used as a critical reader.

A tutor can go further by deliberately presenting the student with a plausible but flawed AI interpretation and asking them to dismantle it using the text. Spotting weak analysis is itself an analytical skill.

IB: From Guided Analysis to Independence

The same principle works particularly well with IB English Language and Literature.

In IB Paper 1, students analyse previously unseen texts. A guiding question provides a possible route into the text, and students are expected to examine how authorial choices shape meaning.

Imagine a Paper 1 advertisement accompanied by a representative question such as:

How do visual and linguistic features work together to influence the intended audience?

Initially, a tutor might allow the student to analyse the advertisement and then compare their observations with AI-generated analysis.

Next time, AI might only be permitted to ask questions:

Who is the intended audience?

What assumptions does the advertisement make about them?

Why has this image been positioned here?

What changes if we interpret this phrase ironically?

Eventually, the AI disappears altogether, and the student completes a timed unseen analysis independently.

That final stage is crucial.

Gradually Remove the Scaffolding

This approach comes from a simple educational principle that informed my postgraduate work on AI and academic writing: support should help learners reach independence, not become permanently attached to the task.

A sensible progression might look like this:

Modelling → guided practice with AI → student writing → feedback → reduced AI assistance → timed independent writing.

AI can play several useful roles along the way. It can provide an alternative interpretation, question a thesis, identify gaps in reasoning, generate a fresh practice question or act as a critical reader of a completed paragraph.

What it should not routinely do is perform the intellectual work the student is supposed to be learning.

This also gives tutors an important role. Rather than attempting to police AI use - an increasingly difficult exercise - tutors can teach students how to use it critically and when to stop using it.

The Examination-Room Test

For parents, there is a simple way of judging whether AI is helping.

Do not ask whether it has made today's homework better. Ask whether the student is becoming more capable of doing tomorrow's work without it.

If a student moves from needing AI to construct an argument, to using it only for feedback, and eventually to producing a strong timed response independently, the technology has served a genuine educational purpose.

If removing AI causes the quality of the student's work to collapse, it has probably been overly relied on.