AI Tools for DSE Revision: Which Ones Are Worth Using and How to Use Them Honestly

S5 and S6 students are, almost without exception, already using AI tools for DSE revision. The question isn't whether they should — that ship has sailed — but whether they're using them in ways that actually help them learn, and whether they can sleep at night knowing they're being honest about it.

This is my practical guide: specific tools, specific use cases, and the line I'd draw between use that supports learning and use that undermines it.

The fundamental principle

Any AI tool use that increases your understanding is legitimate. Any AI tool use that produces output you submit or apply without understanding is not.

This sounds simple but requires thought in practice. Using ChatGPT to explain why a particular answer structure is preferred in DSE history — and then writing your own response in light of that understanding — is legitimate. Using ChatGPT to write the response and submitting it as your answer is not. The distinction is whether you're the one doing the thinking or the AI is.

For self-study and revision — where what matters is whether you actually learn, not what's submitted — the principle becomes: does this tool help me understand more, or does it substitute for the understanding I need to develop?

Genuinely useful applications by subject type

For all written subjects (Chinese, English, History, Economics, Business):

AI as a critic is one of the most underused revision techniques. Write your response to a past paper question. Then ask ChatGPT (or a similar tool) to critique it — specifically asking: "What arguments did I miss? What evidence could strengthen my analysis? Where is my reasoning weak?" This use treats AI as a thinking partner that pushes back, not a machine that produces answers.

The output of this process is your improved thinking. The AI didn't write your essay — you did. It just told you where you fell short, the way a good teacher would.

For Maths (Core and Extended):

Photomath's step-by-step breakdowns are useful for understanding why a method works, not just confirming that an answer is correct. For topics where you keep making the same mistake, use AI to explain the underlying concept from multiple angles until one clicks. Asking "explain this as if I've never seen it before" sometimes produces a different entry point that resolves a persistent confusion.

AI cannot reliably replace doing problems, but it can replace textbook explanations that haven't worked.

For Science subjects (Biology, Chemistry, Physics):

AI is useful for explaining mechanisms at different levels of depth. "Explain enzyme inhibition for a student who understands basic chemistry" gives a different explanation than asking without context. When you're stuck on a concept, asking for different analogies can be effective.

Critical note: verify factual claims in scientific topics independently. AI makes confident errors in chemistry equations and biological pathways. Never rely on an AI explanation of a mechanism without cross-checking against your textbook or past-paper mark schemes.

For Liberal Studies (now Citizenship & Social Development) and any essay-heavy paper:

Brainstorming essay structures and arguments with AI is legitimate. "What are five perspectives on this social issue?" gives you raw material to evaluate, accept, reject, and extend. You're not using the AI's argument — you're using it to surface considerations you can then assess yourself.

Do not use AI to write your practice essays for CSD. The entire point of practice essay writing is developing the capability to produce arguments under time pressure. That capability only develops by doing it yourself.

DSE-specific limitations to know

AI tools are trained on general text data and have uneven knowledge of HKDSE-specific requirements. Mark schemes, assessment criteria, and marker preferences for specific DSE papers are not reliably represented in AI training data.

This means AI feedback on "does this answer look good?" is unreliable for DSE purposes. Your Chinese essay may be linguistically competent but structured in a way the DSE examiner would penalise. An AI tool without DSE mark scheme access cannot know this.

For DSE-specific feedback — what markers actually reward and penalise — past paper mark schemes, examiner reports, and teachers who mark DSE papers are irreplaceable.

How to be honest about what you've used

If your school's AI policy requires disclosure, follow it. But beyond compliance, build the habit of honesty with yourself. Keep a simple mental test: if your teacher asked you to explain your work, could you? Could you extend it? Could you write something similar independently under exam conditions?

If yes: you've used AI as a tool while doing the learning yourself. That's legitimate and increasingly normal.

If no: you've borrowed output without acquiring the underlying understanding. That shortcut will show in your actual exam performance, regardless of what your practice work looks like.

The DSE is still sat in a room, on paper, without AI access. The learning has to happen in your head eventually. AI tools can help it get there — or help you avoid getting there. The choice is genuinely yours.