Six Cross-Subject AI Lessons for Secondary Schools: From Chinese Writing to Science Inquiry

Last updated: 2026-07-22

The commitment of at least three subjects, two year levels each, means a secondary school cannot leave AI sitting in the ICT room. The six cross-subject lessons below cover the languages, maths and science, and the humanities. Every one of them can grow into a teaching demonstration the programme asks for.

1. Chinese: the writing feedback loop

An AI marking tool turns one-shot marking into several rounds of revision. Students hand in a first draft, the AI gives an item-by-item first mark against the rubric along with point-by-point comments, students revise and resubmit, and the teacher reviews the final draft and decides the mark. For senior forms, choose a tool with the HKDSE marking scheme built in so the comments track the real assessment. Junior forms can lean towards content and structure feedback. The teacher's marking time shifts from close marking of every script to reviewing and following up, while students revise far more often. This is the kind of change that is easiest to put into numbers in a report.

2. English: speaking practice for the exam

The bottleneck in speaking practice is the teacher-to-student ratio. One teacher cannot listen to thirty students rehearse every week. In the classroom: students use an AI voice tool to run mock individual responses or group discussions, and the tool records fluency and word choice on the spot. Class time then goes to the teacher modelling fixes for shared weaknesses and running live group practice. The AI handles the volume of practice, the teacher handles the quality.

3. Maths: diagnosing the working, not the answer

Have students submit their working rather than their answer for the AI to check, and ask it to point out the first step that goes wrong and explain the concept, without giving the correct solution. Students fix it and resubmit until the working holds together. From the back end or the chat log, the teacher sees where the class gets stuck most often and teaches to that in the next lesson. One setting matters: the AI diagnoses and does not answer for the student, and this has to be locked in the prompt or the tool's settings.

4. Science: comparing experiment data with AI

After a practical, students first analyse the data and draw their own conclusion, then put the same question to the AI and compare the two. Does the AI's explanation go beyond what the data supports? Does it ignore experimental error? This sequence, yourself first, then the AI, then a comparison, trains scientific reasoning and AI literacy at the same time. Students see for themselves that the AI can be fluent without being rigorous.

5. History and the humanities: a source-checking workshop

The teacher uses AI to generate a realistic but flawed piece of historical material, with wrong dates or one figure's deeds attributed to another, and mixes it in with genuine sources. In groups, students check it against their textbook and reliable sources, find the fake and set out their evidence. A harder version lets students generate their own fakes with AI to test their classmates. Whoever sets the question has to be better at verifying than whoever answers it. Telling good sources from bad has always been a core skill of the humanities.

6. Citizenship and Social Development: a structured debate on AI ethics

Run a structured discussion around real situations: deepfake scam stories, where the line sits on AI-generated homework, algorithmic bias. Students first gather arguments for and against with AI, then have to back each argument with a source that is not the AI, and finally debate a position. The rules make the AI a starting point rather than the last word: "the AI said so" does not count as an argument. That single classroom rule is already a lesson in literacy.

Turning the six examples into evidence for your report

The six examples span six subjects, more than the at-least-three requirement. In practice you can start with three or four subjects the panels are willing to take on, each across two year levels. For every example, record the same things: topic and objective, the AI's role, the lesson flow, samples of student work, quantitative measures such as number of revisions, volume of practice and participation rate, and the teacher's reflection. Design the format once and use it across the school. When you write the mid-term report in 2027, the teaching-demonstration section is just a matter of pulling these together.

Common questions

Senior form time is tight. Will AI teaching slow the syllabus down?

Pick only the parts that are already slow: essay marking, speaking practice and diagnosing working are all bottlenecks in what you already do. AI raises the volume of practice and the number of feedback rounds within the same lesson time. It does not add lessons. If any AI step slows the class down, adjust it or drop it.

What about students using AI to cheat on homework?

Blocking it does not work, and redesign beats blocking. Move the weight of assessment onto the process, such as drafts, revision history and oral defence, instead of a single finished product, and set out clearly where AI may and may not be used. The classroom rule in example six, AI as a starting point but not as an argument, can be spread across the whole school.

Sources

EDB Circular Memorandum 221/2025 (Chinese)

Want to know your school's next step?

Tell us where your school is with the funding programme. We help with evidence, from choosing tools to filing the reports.

Book a free school consultation Request a proposal