Teaching approach

The focus is on writing as thinking and as core to how students develop. When students struggle to choose words, connect evidence, organise paragraphs and explain claims, they are developing understanding.

The approach here is to support that struggle, not remove it. The toolkit is heavily oriented towards students bringing an attempt, a draft, a paragraph, a plan or a question they have already started to form.

Broadly, the steps are to diagnose issues with the student’s work, explain the issue in plain English so the student can understand why it matters, give made-up examples of similar problems, and then prompt the student to apply the idea themselves with further follow-up. The toolkit may provide some fixes to actual texts where the student is truly stuck, but this is limited on purpose.

Ideas first

Clear writing is not only shorter or simpler writing. It must also say exactly what the student means. This principle draws on Orwell’s concern with vague or inflated language, Williams’ focus on clear characters and actions, and plain English principles of reader-centred accuracy. If a change may alter the meaning, the tutor should explain the difference and ask the student to choose.

The toolkit especially discourages “academicising” that swaps a student’s key terms for smoother but different terms. AI use can produce an avalanche of polished academic-sounding language, but that polish can flatten the student’s voice and hide what they actually mean. The approach here is to make the student’s own meaning clearer, not replace it with smoother language.

For paragraph work, the toolkit asks students to connect the chain of ideas before polishing the topic sentence. The aim is to make the paragraph mean more clearly, not just sound smoother. A paragraph may need better wording, but wording is not the first question. The first question is whether the reader can follow the claim, the evidence, the explanation and the link between them. If the central claim or chain of meaning is unclear, polish can hide the problem rather than solve it.

Manageable feedback

The toolkit asks AI to write in short, readable paragraphs by default. Bullet points, tables and lists are still useful when they make feedback easier to act on, but the default should feel more like a focused tutor conversation than a long report.

The aim is to give students a manageable amount of feedback, usually with one clear next action rather than a catalogue of every possible issue.

This matters for tutoring. Students do not only need information they can scan. They often need a short explanation they can actually read, follow and think with. A good paragraph already supports both skimming and understanding. The first sentence gives the reader the main point. The rest of the paragraph explains, illustrates or qualifies that point.

Dealing with “I’m stuck”

Students should be able to say “I’m stuck” at any point. If the reason is clear, the tutor should name the likely problem gently and offer help with that. If the reason is unclear, it should ask a short clarifying question. The response should usually give two or three manageable ways forward in short paragraphs, then ask whether one of them fits or whether the problem is somewhere else.

Development, not completion

AI tools optimise for completion. That can undermine education by trying to fix writing too quickly. This toolkit is designed to optimise for understanding: helping students see what is wrong, why it matters, and what they can try next.

No tutor should simply provide the answers for a student. It is vital that the AI tutor follows the same principle.