Artificial Intelligence

AI Usage
← Back to guidelines

Working responsibly with AI

AI can support ELT writing when it is used carefully, with human review, clear pedagogical goals, and respect for accuracy, privacy, and copyright. The aim is to improve efficiency and creativity without reducing quality or accountability.

◎Purpose

The aim of using AI in language training material development is to support idea generation, drafting, and editing in a way that saves time and improves workflow. Used well, AI can help writers explore options more quickly, generate starting points, and check language or structure before human review. However, AI should always be used carefully so that pedagogical quality, level control, learner privacy, copyright, and editorial responsibility remain firmly in human hands.

1General Principles

  • Use AI for brainstorming, outlining, and language checking, but keep full responsibility for the final pedagogical design, accuracy, and appropriacy.
  • Treat AI output as a drafting assistant, not as a source of ready-to-publish content.
  • Always adapt, refine, and localise the output before publication.
  • Follow all existing project guidelines, including level, objectives, accessibility, diversity and inclusion, PARSNIPS, and copyright rules.

2Good Uses of AI

  • Brainstorming and generating ideas such as topics, scenarios, role-plays, and contexts that fit the brief.
  • Producing rough first drafts of short texts, emails, dialogues, announcements, and similar items that can then be edited for level, style, and target language.
  • Suggesting distractors, paraphrases, and reformulations for comprehension questions or vocabulary work, with careful checking for level and meaning.
  • Helping to simplify or grade existing texts while preserving key lexical items, task purpose, and overall coherence.
  • Drafting rubrics or instruction options that can then be rewritten in house style.
  • Supporting consistency checks such as terminology, spelling variety, and key term usage.

3Prohibited or High-risk Uses

  • Do not submit content that is largely or entirely AI-generated without sufficient adaptation and editing.
  • Do not use AI to imitate or adapt copyrighted material.
  • Do not use AI output as a factual source for statistics, legal claims, medical claims, company details, or similar information without independent verification.
  • Do not paste confidential learner, client, or company data into AI tools.
  • If a brief requires real-world data, use anonymised or publicly available information instead.

4Quality and Level Control

  • Check CEFR level and remove or replace language that is too advanced, too rare, or not in scope.
  • Make sure function, register, and task type match the brief exactly.
  • Edit for naturalness so the final text sounds like a real teaching material rather than an AI draft.
  • Align the content with the intended learning outcomes, not simply with what AI produces first.
  • Run a final pedagogic sanity check for clarity, noticing, practice, feedback, and any cultural or diversity issues.

5Transparency and Documentation

  • Keep a simple note in the project files when AI has been used.
  • Flag any AI-generated content that needs editorial review because of possible pragmatic, cultural, or factual issues.
  • If in doubt about whether AI use is acceptable for a specific project, consult the project lead before proceeding.

6Prompting

If you create bad prompts, you will create bad output. Follow an established prompting method, such as CREATE or RISEN, and be specific about the audience, task, output type, and constraints.

CREATE Framework
  • Character: define the role you want the AI to play.
  • Request: state the task clearly.
  • Examples: give models of the output you want.
  • Adjustments: explain what should be refined or changed.
  • Type of output: specify the format, such as table, bullet points, or continuous text.
  • Extras: add useful details or constraints.
RISEN Framework
  • Role: establish the AI’s role and perspective.
  • Instructions: give a clear directive.
  • Steps: break the task into manageable stages.
  • End goal / expectations: define the desired result.
  • Narrowing / novelty: add constraints or features that make the response more focused and useful.

✓Key Takeaway

  • Helpful: use AI to speed up idea generation and drafting.
  • Careful: check everything for level, accuracy, and appropriacy.
  • Transparent: keep a record of where AI was used and how the output was edited.
Example of responsible use
Use AI to draft three role-play scenarios, then select the best one and rewrite it for the target level.