Digital & AI English · Secondary 2 English · English Atlas
AI can make English work faster. That is not the same as making the student better at English.
The central tutoring decision is therefore not whether AI is good or bad. It is which jobs AI may do without removing the cognitive work the learner is supposed to acquire. For Secondary 2, the safest rule is simple: AI may support the learning process, but the student must still own the reading, reasoning, language decisions, authorship and final independent performance.
The short answer
- AI may ask questions.
- AI may offer alternative explanations.
- AI may generate additional practice after the target is defined.
- AI may compare versions and help the learner notice differences.
- AI may suggest checks for grammar, evidence, audience and scope.
- AI may simulate discussion and follow-up questions.
- AI should not replace the learner’s first reading, first reasoning, first plan or final answer when those are the abilities being trained.
- AI-assisted success should always be followed by an AI-off transfer test.
1. Start from the learning job
Before opening an AI tool, ask: what human capability is this task supposed to build? If the goal is inference, the learner must interpret evidence. If the goal is paragraph development, the learner must decide what the example proves. If the goal is editing, the learner must learn to notice and repair language patterns. AI should be positioned around that job, not in place of it.
2. Performance and learning are different
The OECD Digital Education Outlook 2026 warns that general-purpose generative AI can improve immediate task performance without producing equivalent learning gains when students offload the cognitive work. That distinction matters enormously in tuition. A better-looking paragraph is not automatically evidence of a better writer.
3. AI can be a questioning partner
One of the strongest uses of AI is to ask the learner questions instead of supplying the answer. For reading: “Which sentence most directly supports that inference?” For writing: “What does your example prove?” For editing: “Which part of the sentence controls agreement?” Good questioning keeps the learner cognitively active.
4. AI can offer a second explanation
If a tutor’s explanation does not land, AI can provide another analogy or worked example. The learner should then explain the concept back in their own words and apply it to fresh material. The second explanation is useful only if it changes understanding rather than merely sounding clear.
5. AI can generate contrast pairs
Contrast is powerful in English learning. Ask AI for two versions of a sentence, paragraph opening or inference—one stronger and one weaker—then require the student to identify the dimension that changes. The learner does the judgement; AI supplies material for judgement.
6. AI can generate fresh practice after diagnosis
If the tutor has already identified a precise target, AI can help create additional examples or practice sets. A learner working on qualifier preservation might receive new sentences containing may, often, unless and only when. Practice generation is useful when the mechanism is already defined and the outputs are checked for quality.
7. AI can support retrieval
AI can quiz the learner without immediately revealing the answer: “List the three checks you use before changing a comprehension answer.” “What are your current writing risks?” “What should happen after feedback?” Retrieval should come before hints.
8. AI can support vocabulary depth
Ask AI to generate contextual comparisons between near-synonyms, collocations or register choices. The learner should choose and justify. A useful interaction might compare cautious, hesitant and reluctant across different sentences. The tool creates variety; the student builds precision.
9. AI can support oral rehearsal
AI can ask follow-up questions that force the student to extend, qualify, disagree or recover. This is valuable when the learner answers aloud before seeing any suggested response. The AI should behave like an interviewer, not a scriptwriter.
10. AI can support feedback comparison
After the learner completes a piece independently, AI feedback can be compared with teacher or tutor feedback. Where do they agree? Where do they disagree? Which comment is supported by the task requirements? This turns AI output into an object of evaluation rather than an authority.
11. AI can support revision choices—but not own the revision
The learner can ask for three possible ways to improve a paragraph, then decide which advice is relevant and rewrite the paragraph personally. If AI rewrites the whole section, the final product may improve while the student loses the chance to practise the actual revision decision.
12. AI should not do the first reading when reading is the target
If the student asks AI to summarise the passage before reading it, they may remove the need to build a text model themselves. For reading tuition, the learner should usually read first, annotate or explain first, and only then use AI to challenge or compare interpretations.
13. AI should not do the first plan when planning is the target
A ready-made essay plan can hide whether the student can interpret the prompt, select ideas and organise them. If planning is the learning job, the first plan must belong to the learner. AI may critique it afterwards or offer a contrasting plan for comparison.
14. AI should not become the student’s sentence engine
When AI generates polished sentences continuously, the student can become an editor of someone else’s language instead of a writer. Secondary 2 tuition should preserve original sentence production. AI can provide comparison examples after the learner has written.
15. AI should not hide uncertainty
Singapore’s Student Learning Space guidance explicitly reminds learners that generative AI responses are probabilistic and may be inaccurate, and encourages verification. This is an important English-learning habit: fluent language is not proof of correctness.
16. AI should not replace assessed authorship
For assessed or submitted work, school rules matter. Some tasks prohibit or restrict AI assistance; others require disclosure. The learner should follow the institution’s rules and preserve authorship. Generating coursework and presenting it as one’s own can create serious academic-integrity problems.
17. Maren: AI as a questioning editor
Maren writes a paragraph independently. Instead of asking AI to rewrite it, she asks for questions that test whether each example has been interpreted. AI asks, “What exactly does this example prove?” Maren revises personally. A later fresh paragraph without AI tests whether the question became internal.
18. Iona: AI as an evidence challenger
Iona answers an inference question first. Then AI is asked to propose an alternative interpretation supported by the text. Iona must compare the evidence and explain which interpretation is safer. AI widens the comparison; Iona retains judgement.
19. Leonie: AI as a recovery simulator
Leonie practises a timed sequence without AI. Afterwards she asks AI to generate three “what if” disruptions: an unfamiliar question, a difficult paragraph opening and a late time warning. She explains her recovery decision for each. The tool expands scenarios without taking over execution.
20. Use an AI job card
- Learning target: what human capability are we training?
- Student-first action: what must the learner attempt before AI?
- AI job: question, explain, compare, generate practice or verify?
- Verification: how will AI output be checked?
- AI-off test: what fresh task will prove independent control?
21. Green, amber and red AI jobs
Green: asking follow-up questions, generating extra examples, creating contrast pairs, quizzing retrieval, simulating oral discussion. Amber: giving feedback, suggesting revisions, summarising difficult concepts—use only after an independent attempt and with verification. Red: producing the learner’s assessed answer, replacing first reading or first planning when those are the target skills, or creating work that is submitted as the student’s own.
22. Where this page fits
This page owns the Secondary 2 tutoring decision about AI’s permitted jobs. It does not replace broader AI-literacy owners elsewhere in the eduKate estate. For the standard tutoring lesson cycle, use What Should Happen in an English Tuition Lesson?. For transfer to school work, use How to Make English Tuition Transfer to School English.
23. Further reading
- OECD Digital Education Outlook 2026
- Singapore Student Learning Space: Guidance on Generative AI
- Ofqual: AI in coursework resources for schools
24. Final test
A good AI job leaves the student doing more meaningful thinking, not less. The strongest sign is simple: after AI disappears, the learner can still perform the English decision independently.
Explore the Secondary 2 English hub for related guides.