● Under constructionAll modules are still being built — expect rough edges and missing content.
Deeply Materials.

Using AI at the Language Centre

Using AI at the Language Centre.

Here are nine practical ways to use AI. Each one includes a prompt you can copy, practical advice and a clear explanation of what can go wrong. First, read the rules about what information you may enter into each tool, as they apply throughout the guide. You should also read the five habits at the end, which are more important than any individual prompt.

Before you paste anything

Closed by default – but read this before you paste anything into an AI tool. It determines what data you are allowed to enter into each one.

Preparing to teach

Section 01

This section covers the preparation you do before a class, including drafting materials, planning the session and building enough background knowledge on an unfamiliar topic.

Materials and texts

Section 02

This section covers work on documents you already have, as well as the production of new material to use with students.

Assessment and student work

Section 03

This section covers feedback on student writing, assessment criteria and the question of how students themselves may use AI.

Getting better at this

Five habits

Prompts copied from a page will only take you so far. These five habits will improve your results more than any individual prompt, and they apply to anything you ask an AI to do.

  1. Ask the model to interview you first

    Begin by writing “before you draft anything, ask me three questions about my group and my constraints”. The model's assumptions about your students are almost always wrong, and this is the simplest way to correct them.

  2. Show the model your own materials

    Paste in an exercise you already use and ask for more of the same on a different topic. A single example of your own style shapes the output far more effectively than any description of that style in words.

  3. Treat the first answer as a draft

    Every prompt on this page is the beginning of a conversation rather than a single request. Tell the model that the tone is too formal, that the third point needs redoing, or that an example will not suit engineering students. The version you can actually use usually appears after three or four exchanges.

  4. Save the instructions you keep retyping

    If you describe your course, level and students every time you start, store that description somewhere you can reuse it, either as a CurreChat background prompt attached to the course or as saved custom instructions. On a course you teach every year, this quickly saves time.

  5. Decide in advance what you will check

    Anything you would otherwise have had to look up is what the model is most likely to get wrong while sounding certain, including figures, dates, citations, claims about CEFR levels and word counts. Decide which parts of the output you intend to verify before you ask for it.

CurreChat and Copilot, signed in with your university account, are the tools the University provides, and you should check what data each one accepts before you paste anything into it. Always review AI output before you share it with students, as it works best as a first draft rather than as a finished product. Assessment decisions remain yours, and AI does not grade student work. This page was last reviewed in August 2026.

This is an unofficial guide, not a University publication. It reflects tools, policies and data-protection rules as they stood at the time of writing, and no responsibility is accepted for errors, omissions or changes since. AI tools and University policy both move quickly, so treat anything here about what is currently permitted as a starting point, and check the University’s own pages for the current position: Generative AI at the University and AI in teaching.