The AI Guide for Language Educators

From Core Concepts to Advanced AI-Enhanced Pedagogy

AI Guide

Welcome to Your New Co-Pilot

The world of AI in education can feel exciting, and maybe a little overwhelming. This guide is designed to demystify generative AI and empower you to use it as a powerful teaching assistant. Think of AI not as a replacement, but as a "co-pilot" that handles tedious prep work, so you have more time to do what you do best: inspire and connect with your students.

The Productivity Trap: Avoiding AI Burnout

AI is meant to save you time, not double your workload. Keep these boundaries in mind to protect your energy:

  • Beware "Expectation Inflation": Just because AI can generate five different reading levels in seconds doesn't mean you must do it for every class. Use the time you save to catch your breath, not to overproduce.
  • Invest Time to Save Time: Learning to prompt effectively is a new skill. Avoid overwhelm by starting small, automate just one task (like brainstorming warm-ups) before trying to change your whole routine.
  • Quality Over Quantity: A flood of AI-generated content exhausts both you and your students. Focus on using AI to elevate a few core materials rather than churning out endless worksheets.
  • Protect Your Off-Hours: AI works 24/7, but you shouldn't. Use your newfound efficiency to confidently close your laptop and step away.
To understand why AI might actually be increasing your workload instead of reducing it: Read the HBR Article

Audit Your Workflow: What should you actually automate?

Go to JobsGPT (a tool to assess AI's impact on tasks and workflows) and search for the specific role of an English Language Teacher. Explore the results to discover for yourself which tasks are highly exposed to automation and which areas require irreplaceable human connection.

💡 Key Takeaway: The ELT profession is evolving toward a hybrid model. Human-centered functions remain highly resistant to automation. Let AI handle the routine preparation and personalization, so you can act as a learning coach, facilitator, and cultural communicator.

When Copilot Becomes Autopilot: Cognitive Offloading

The greatest risk of Generative AI isn't factual errors—it's the atrophy of our critical thinking.

It is tempting to let AI do all the heavy lifting, a phenomenon known as cognitive offloading. We naturally tend to "satisfice" (accept solutions that are merely "good enough"). When we rely on AI to generate our ideas, lesson plans, and content without deep reflection, we risk going on "autopilot," outsourcing our critical judgment to a machine.

The Danger for Educators

When teachers consistently offload instructional design, it leads to a mechanised convergence effect—meaning materials become homogenous, predictable, and lack diversity. We risk deskilling ourselves and losing the deep, personal connection to the pedagogy we design.

The Danger for Students

Learning requires struggle. When students use AI to bypass the effort of formulating thoughts or finding the right words, they eliminate the essential "friction of learning." Bypassing this necessary cognitive struggle directly short-circuits true language acquisition.

The Solution: AI as a "Provocateur"

Instead of using AI merely as an obedient assistant to generate answers, design your workflows to challenge your thinking. Ask the AI to critique your lesson plan, highlight biases, play devil's advocate, or suggest entirely contrasting perspectives. Engage in critical integration where AI provokes reflection, but you remain the ultimate decision-maker.

Based on the research paper "When Copilot Becomes Autopilot Advait Sarkar.pdf". Read the Full Paper

What is Generative AI?

Understanding what AI is (and isn't) is the first step to using it effectively.

The easiest way to think of Generative AI is as a super-powered autocomplete. It has been trained on a vast library of text and images, and its main skill is recognizing patterns to predict the next most likely word or pixel.

✅ It is a Creation Engine

GenAI excels at generating new, original content. It can write poems, create lesson plans, or produce dialogues that have never existed before. It's a creative partner for brainstorming and material creation.

❌ It is NOT a Search Engine

AI does not "know" facts or "understand" concepts like a human. Because it only predicts the next word, it can be confidently wrong and even make up information (this is called a "hallucination"). Always fact-check its output.

How AI Reads: RAG & Chunking

When you upload a large PDF, the AI doesn't read it cover-to-cover like a human. It uses a process called RAG (Retrieval-Augmented Generation). It breaks the document into "chunks" and searches for the pieces most relevant to your prompt. Because of this, it might omit information or provide out-of-context answers. Always double-check its summaries against the original document.

Teaching ABOUT AI vs. Teaching WITH AI

A core MEC and UNESCO guideline is this crucial dual approach. While we use AI as a tool for learning (Teaching WITH AI), we must simultaneously teach students how AI works—its data, algorithms, and ethical implications (Teaching ABOUT AI). Students need to understand the tool they are using to remain critical thinkers. Furthermore, the Human-in-the-loop principle dictates that AI is an assistant, not a replacement. The educator must always remain the central decision-maker and mediator.

Getting Started: A Key Pairing for Beginners

New to AI prompting? Start with this simple yet powerful combination to get immediate, high-quality results.

Pairing Suggestion:

Combine Instruction-Based Prompting (#4) with a Role/Persona Prompt (#14).

Example:

"You are a friendly, encouraging language tutor. Summarize this B2 news article for B1 learners in 3 bullet points using simple language."

Why it works:

This pairing tells the AI what to do (the instruction) and how to be (the persona), leading to outputs that are not only accurate but also have the right tone and pedagogical focus.

Core Principles of Effective Prompting

Before exploring specific techniques, master these universal principles to boost the effectiveness of every prompt you write.

Remember the PARTS Acronym

Structure your prompt using PARTS to cover all key elements.

  • Persona: The role the AI should play.
  • Aim: Your objective or the problem to solve.
  • Recipients: The audience for the response.
  • Theme: The desired style, tone, and register.
  • Structure: The desired output format.

Be WISE with AI

Use this acronym as a checklist for critical and responsible use.

  • Watch out for hallucinations and bias.
  • Inspire creativity, don't replace it.
  • Safety first; protect personal data.
  • Ethical use; respect originality and avoid plagiarism.

Dos and Don'ts

Do

  • Assign a role & expertise level.
  • Be specific about the output format, length, and tone.
  • Provide examples of good output when possible.
  • Ask for step-by-step reasoning for clarity.
  • Set constraints and ask for checks or alternatives.
  • Iterate and refine your prompt if the first result isn't perfect.

Don't

  • Ask for an internal monologue (e.g., "show your chain-of-thought").
  • Be vague about the output format you want.
  • Forget to provide the core problem or data to be analyzed.
  • Use it as a substitute for professional medical, legal, or financial advice.

Stuck? Use a Prompt Generator

If you are struggling to write the perfect prompt from scratch, prompt generators can do the heavy lifting for you. Simply fill in a few fields about what you need, and these tools will construct a highly optimized prompt that you can copy and paste directly into your AI.

  • Prompt Cowboy Generate optimized prompts easily. (Note: The free tier offers a limited selection of features).
  • Feedough A robust AI prompt generator offering a wider range of features completely for free.

Or... Ask the AI to Build the Prompt for You

Instead of using external generators, you can use the LLM itself to craft your prompt. Provide the AI with your raw ideas using the PARTS acronym, and ask it to build the optimized prompt.

TRY THIS META-PROMPT:

"I need to create a prompt for a task. Here is the raw information based on the PARTS framework: [Insert your Persona, Aim, Recipients, Theme, Structure]. Please generate a highly optimized prompt that I can use to achieve this aim. I will analyze your suggested prompt and we can refine it together."

Important: Always analyze the result the AI gives you, rewrite or tweak it if necessary, and then copy it to use in a fresh chat.

The Master Prompt & Context Window

Unlock the true power of AI by setting a strong foundation and understanding its memory limits.

1. The Master Prompt

A Master Prompt is a foundational document that contains your educator profile, teaching context, student demographics, and pedagogical goals. Instead of pasting this at the beginning of every chat, you can add this to your AI's settings. By establishing this context upfront, you ensure highly tailored, relevant responses.

How to Save Your Master Prompt

Most major AI models allow you to save these instructions so they apply to all future chats:

  • Google Gemini: Go to Settings > Personal Intelligence > Instructions for Gemini > Add.
  • ChatGPT: Go to Settings > Custom Instructions.
  • Claude: Use "Projects" or set custom instructions depending on your subscription tier.
Template: "I am an English teacher at a [public/private] school in [City]. I teach [age group] at the [CEFR Level]. My pedagogical approach is based on [TBLT/Communicative Approach]. Please tailor all activities to be highly interactive, culturally relevant to Brazil, and require minimal printing."

How to Build Your Master Prompt from Scratch

Don't want to use a template? Ask the AI to interview you to create your perfect Master Prompt. Open a new chat and follow these two steps:

Step 1: The Interview Prompt

Copy and paste this into the AI:

"I want to create a professional profile to use in generative AI tools. Ask me questions about who I am, my professional role, my work context, my areas of interest, my current projects, my professional goals, my communication style, and how I like to receive guidance."
Step 2: The Synthesis Prompt

After you answer the AI's questions, reply with this:

"[Paste your answers here]. Based on my answers, synthesize everything into a document called 'MASTER PROMPT FOR AI', ready to be used on different AI platforms so I receive better responses. Summarize everything in up to 1200 characters."
What NOT to Include in Your Master Prompt
  • Passwords
  • Full address
  • Document numbers (ID, CPF, SSN, etc.)
  • Bank details
  • Intimate or deeply personal information
  • Information about other people (students, colleagues) without authorization

2. Context Window vs. Memory

AI doesn't "remember" past conversations like a human. It has a Context Window—a limited amount of text (tokens) it can keep active in its "short-term memory" at one time. In a long conversation, as you add new prompts, the AI will eventually "forget" your Master Prompt and initial instructions, even if they are saved in your settings.

🐦 The "Canary" Analogy (Testing the Context Window)

Miners used to take canaries into coal mines as an early warning system. You can use a clever trick to build a similar warning system for your AI!

Add this instruction to your Master Prompt in your AI settings:
"Always start every response to me with the word 'Canary'."

If the AI gives you a response that does not start with "Canary", it means its context window is full. It has "forgotten" your initial instructions, and it is time to start a brand new chat!

Using AI with Students

Establish clear guidelines for safe, effective, and ethical use of AI in your classroom.

Paradigm Shift

AI detection tools are often flawed and biased against non-native speakers.

Instead of policing, talk to students about the use of AI and teach them how to use it in a way that develops their critical thinking.

Explain what is and what is not okay to do with AI. It might seem obvious to you, but not to them.

Reflect and ask students: How did they interact with the AI? Did they accept poor suggestions or use critical thinking to refine the text?

Rule #1: The Teacher is the Pilot

AI use must always be supervised and purposeful. Before any activity, explain the "why" to students: "We are using this tool today specifically to help us brainstorm ideas," or "...to practice vocabulary."

Rule #2: The Digital Safety Briefing

CRITICAL: Data Protection (LGPD & ECA Digital)

Strict rules exist against sharing sensitive student data. Never input real student names, grades, behavioral reports, or psychological diagnoses into AI models. AI platforms may use your inputs for training, causing severe privacy violations. Always use fictionalized personas or fully anonymized data when prompting.

  • It's a Tool, Not a Friend: Remind students they are interacting with a program, not a person. It doesn’t have feelings.
  • Protect Your Privacy: Treat chatbots like a public postcard. Never share personal details (full name, school, address, photos).
  • Verify Visuals (Reverse Search): Teach students to "flip the search" by uploading an image (using tools like Google Lens) to find its original, authentic source. This is an essential safeguard to help them spot AI-generated fakes, fact-check viral photos, and avoid visual misinformation.
  • Fact-Check Everything: Reinforce that AI can be confidently wrong ("hallucinate"). Students are responsible for verifying all information.

AI for Inclusive Education & Diversity

Technology should be a bridge, not a barrier. In this section, we explore how AI can support neurodiversity and expose students to global varieties of English.

A. Supporting Neurodiversity (Differentiation)

Students with dyslexia, ADHD, or other specific needs can benefit immensely from personalization via AI.

Text Adaptation

Use AI to rewrite complex texts keeping the mature theme but simplifying the syntax.

"You are an experienced educational content writer. I want you to rewrite an article about climate change in a way that engages a Brazilian teen student, aged 13 years old, who has ADHD and B1 (intermediate) English proficiency, ensuring that the content is easy to read and understand while maintaining a mature tone."

Multimodality

Convert text to audio (TTS) for auditory support or ask AI to describe images for students with low vision.

B. Global Englishes & Linguistic Variation

Most AI models standardize English to American or British variants. As teachers, we must break this pattern and expose students to English as a Lingua Franca (ELF).

  • Linguistic Representation & Algorithmic Bias: AI models are trained predominantly on data from the Global North. Regional variants, indigenous languages, and creoles are largely absent. Uncritical adoption of AI can lead to linguistic imperialism and cultural erasure, reinforcing societal biases. Discussing these biases with students is a vital part of AI literacy.
  • Critical Comparison: Ask AI to rewrite a formal text in teenage slang or regional dialect and ask students to analyze the register shifts.
  • Simulating Varieties: Ask AI to generate dialogues in "Singlish," "Indian English," or other variants to train sociolinguistic awareness. However, bear in mind that since most LLMs (like GPT or Gemini) are primarily trained on standardized English, the generated texts in these varieties must be reviewed thoroughly. This presents an excellent opportunity for students to exercise critical thinking.

C. The Pragmatics Lab

We often fail to teach how to use language socially.

Speech Acts

Ask AI to generate politeness variations for class discussion.

"Generate 5 ways to refuse a wedding invitation, ranging from very rude to extremely polite/formal."

Accessible Textbooks

AI isn't just about efficiency. It's a powerful engine for equity. TheAccessible Digital Textbooks (ADT) initiative leverages AI to create universally designed learning materials for children with disabilities across the globe.

Why it matters for ELT: This initiative proves that adopting inclusive AI tools does more than accommodate individual needs. It fundamentally re-skills teachers in inclusive pedagogy. By utilizing AI to generate accessible formats (like audio, simplified text, and visual descriptions) that work offline, language educators can ensure equitable access to education while actively bridging the digital divide.

Discover how AI is driving systemic inclusive education: Read the UNICEF Interview

E. Unplugged AI & Digital Equity

Teaching AI concepts does not always require high-speed internet or state-of-the-art devices. To ensure digital equity in low-tech environments, we can and should use "Unplugged AI."

  • What is it? It involves teaching the foundational logic behind AI (algorithms, pattern recognition, decision trees, and computational thinking) using physical, non-digital activities (e.g., pen and paper exercises, board games, or physical sorting tasks).
  • Why it matters: It democratizes AI literacy. Students in under-resourced schools can still develop critical computational thinking and understand how algorithms shape their world, preventing the digital divide from widening while ensuring no student is left behind.

Curated Resources: The Pedagogical Foundation

Artificial Intelligence in ELT should not replace pedagogy, but serve it. Below, I have selected essential works, authors, and channels that connect technology to SLA (Second Language Acquisition) and TBLT principles.

Official Brazilian Guidelines (MEC)

Essential reading for educators and administrators aligning their practices with national standards for responsible AI use in Brazil.

  • Referencial de Inteligência Artificial na Educação

    (Framework for the Development and Responsible Use of Artificial Intelligence in Education). The official MEC framework detailing pathways for ethical AI integration across all educational levels.

  • Inteligência Artificial na Educação Básica

    A guiding document specifically for K-12 (Educação Básica) focusing on curriculum pathways, critical media literacy, and ethical practices in schools. It emphasizes the dual approach of teaching about AI and teaching with AI.

Recommended Books

Focus: AI in Language Teaching

Researchers to Follow

  • •
    Nicky Hockly: Authority in EdTech and critical digital literacy. [View profile]
  • •
    Stephen Krashen: His theories of Comprehensible Input and Affective Filter are the basis for creating AI prompts that generate leveled texts. [Official Site]
  • •
    Leo van Lier: Proponent of the ecological approach ("pedagogy first, technology last").
  • •
    Benjamin Luke Moorhouse: Actively researches teacher digital competence in the AI era. [Academic Profile]

Instagram Profiles

YouTube Channels

For Teachers

  • Russell Stannard

    Step-by-step tutorials on NotebookLM and Gemini.

  • Teaching English with Oxford

    High-level webinars on the future of the profession.

  • ELT in Brazil

    High-quality teacher training and development for teachers in Brazil and abroad.

  • BRAZ-TESOL

    Official channel for Brazil's largest association of English teachers. Affiliate of TESOL International and IATEFL member.

For Students

Technical Skills

  • Jeff Su

    Learn how to leverage AI tools to streamline tasks, make better decisions, and free up your time to do more of what truly matters.

  • Tina Huang

    Ex-Meta Data Scientist. AI for productivity and efficient learning.

Explore AI Chatbots & LLMs

Essential AI Tools for Educators

Creative & Media

Writing & Language

Education & Learning

Prompting & Planning

Interactive & Data

Further Learning & Resources

Continue your journey into AI-enhanced education with these excellent resources.

You Don't Need Endless Courses to Use AI!

However, if you're interested in diving deeper, here are some excellent suggestions. For educators new to AI, we recommend following this path from foundational concepts to specialized applications in language teaching.

Step 1: Master the Fundamentals (AI & Prompting Basics)

Step 2: Apply AI to Education (General Pedagogy)

Step 3: Specialize in Language Teaching (ELT Focus)

The Future of AI: Agentic Systems

Agentic AI (or "autonomous AI assistants") represents the next major evolution in how we interact with artificial intelligence.

Proceed with Caution

While this technology is incredibly powerful, it is still in its early stages of development. Extreme caution and oversight are required when granting these systems access to your data or digital tools.

Generative AI vs. Agentic AI

The main difference lies in autonomy and the capacity for action.

Generative AI

You provide a prompt, the AI processes it, and generates a response based on patterns. The interaction ends there. It waits for your next command.

Agentic AI

You define a broader goal (e.g., "Organize the class schedule and send calendar invites to the students"). The AI takes the initiative to execute it.

How an Agent Works

1. Plans

It breaks down your overarching goal into a logical sequence of smaller, manageable steps.

2. Uses Tools

It actively accesses external resources like calendars, email clients, spreadsheets, or APIs to perform the actual work.

3. Reflects & Adjusts

It monitors its own progress. If a step fails or hits a roadblock, it attempts a new approach instead of simply stopping.

The Hidden Cost of AI: Ecology & Ethics

It is easy to view generative AI as a consequence-free convenience, but every prompt we issue carries a physical and ethical price tag.

Generative AI feels like magic happening in the "cloud," but that cloud is anchored to massive data centers. As language educators, we must shift from passive consumption to intentional, critical use.

🌍 Environmental Impact

Training and running Large Language Models consumes staggering amounts of electricity and evaporates millions of gallons of freshwater for cooling. These costs are often outsourced to vulnerable, drought-stricken communities.

⚖️ Ethical Extraction

These models are built upon the uncompensated extraction of human creativity (books, articles, art) and introduce real risks of algorithmic bias and misinformation into our teaching materials.

What This Means for the ELT Classroom

  • Do the "Worth It" Test: Don't use a massive neural network for something a simple web search or dictionary could handle. Weigh the convenience against the energy toll.
  • Protect the Friction of Learning: The cognitive struggle of finding the right word is a vital human process. Outsourcing that friction to AI doesn't just waste energy; it short-circuits language acquisition.
  • Teach Critical AI Literacy: Discuss AI's footprint with your students as a real-world case study in technology and the environment.
To learn more about AI's impact on energy and democracy: Read the Greenpeace Report

Prompt Library

Explore comprehensive prompting strategies from foundational techniques to advanced ELT-specific workflows. Note: You don't need to memorize them all. Simply explore and use what fits your needs.

1. Zero-Shot Prompting

EXAMPLE

“Write five A2-level sentences about last weekend using the past simple.”

USE CASE

Quick activity generation without providing examples.

2. One-Shot Prompting

EXAMPLE

“Here is one example of a correction: Input: ‘He go to school yesterday.’ → Output: ‘He went to school yesterday.’ Now correct this: ‘She don’t like coffee.’”

USE CASE

Error correction or task modeling with a single example.

3. Few-Shot Prompting

EXAMPLE

“Example Q-A pairs for B1 speaking: ‘Describe a memorable trip’ → Provide three model answers of 80–100 words. Now produce a new B1 answer to: ‘Describe a teacher who inspired you.’”

USE CASE

Modeling target output for length, lexical range, and register.

4. Instruction-Based Prompting

EXAMPLE

“Summarize this B2 news article for B1 learners in 3 bullet points using simple language.”

USE CASE

Level-appropriate simplification and task specification.

5. Template / Fill-in-the-Blank

EXAMPLE

“Lesson plan: [Level] [Skill] [Time] [Objectives]. Fill for B1 listening, 45 minutes, topic: travel.”

USE CASE

Generating consistent, structured outputs for planning.

6. Output-Specific Prompting

EXAMPLE

“Provide a JSON list of 10 B1 academic words with fields: word, part_of_speech, CEFR_level, example_sentence.”

USE CASE

Formatting outputs for easy import into tools or apps.

7. Data-to-Text Prompting

EXAMPLE

“Given this JSON of a unit plan, generate a concise, human-readable lesson summary in 2 sentences for the course syllabus.”

USE CASE

Converting structured data into natural language.

8. Step-by-Step / Chain-of-Thought

EXAMPLE

“Explain, as a checklist, how to form the first conditional; then provide 5 A2 examples and 3 CCQs.”

USE CASE

Transparently modeling the form, meaning, and use of a language point.

9. CoT + Verification

EXAMPLE

“Grade this paragraph with B1 writing descriptors. Show the criteria used, assign a band, then verify if examples from the text support the score.”

USE CASE

Reliable, transparent assessment aligned with specific rubrics.

10. Self-Consistency Prompting

EXAMPLE

“Create three different sets of CCQs for the present perfect continuous. Compare and pick the set that most effectively checks meaning and form.”

USE CASE

Improving the quality of questions by evaluating multiple options.

11. Debugging / Error-Analysis

EXAMPLE

“Review this lesson plan for logical gaps, timing inconsistencies, or unclear instructions; propose a fix with a rationale.”

USE CASE

Systematic quality assurance check of materials or activity designs.

12. Evaluation / In-Criteria

EXAMPLE

“Evaluate this unit plan against these criteria: alignment to CEFR, inclusivity, workflow efficiency; score each 1–5.”

USE CASE

Structured, rubric-based review before implementation.

13. Uncertainty / Calibration

EXAMPLE

“Recommend two reading texts for a B2 class, report your confidence (0–100%), and state your assumptions about topic familiarity.”

USE CASE

Making calibrated, honest resource selections and identifying knowledge gaps.

14. Role/Persona Prompting

EXAMPLE

“You are a CELTA-certified teacher. Design a 60-minute lesson on comparatives for A2 adults.”

USE CASE

Generating outputs with tailored pedagogy and an appropriate tone.

15. Reframing / Perspective

EXAMPLE

“Explain phrasal verbs with ‘get’ as if teaching 14-year-old beginners; use relatable examples from their lives (e.g., social media, games).”

USE CASE

Adapting explanations to a specific audience's context and interests.

16. Expert Prompting

EXAMPLE

“Respond as an IELTS examiner. Give Task 2 feedback focusing on Task Response, Coherence and Cohesion, Lexical Resource, and Grammar.”

USE CASE

Generating high-stakes, exam-specific feedback and materials.

17. Multi-Persona Prompting

EXAMPLE

“Act as a teacher and a syllabus designer. The teacher wants more speaking practice; the designer is concerned about curriculum coverage. Discuss a compromise.”

USE CASE

Balancing competing priorities and exploring different viewpoints.

18. Red-Team / Adversarial Check

EXAMPLE

“Challenge this new language school policy for ELLs: identify potential equity risks and implementation gaps; propose mitigations.”

USE CASE

Stress-testing policies or materials for unintended consequences.

19. Decomposition & Execution

EXAMPLE

“First, outline a 4-week B1 speaking module. After I confirm, draft Week 1’s detailed lessons.”

USE CASE

Breaking complex design tasks into manageable, verifiable steps.

20. Constraint-Based Prompting

EXAMPLE

“Create a 50-word A2 dialogue at the supermarket using at least 5 countable nouns and no phrasal verbs.”

USE CASE

Creating controlled practice activities that focus on specific language.

21. Constraint Satisfaction → Optimization

EXAMPLE

“Design a 20-minute A2 speaking task with no printing and pairs only. Then optimize it for maximum engagement.”

USE CASE

Finding creative solutions within real-world classroom limitations.

22. Iterative / Refinement

EXAMPLE

“Rewrite these task instructions for A2 learners. Now make them even more concise.”

USE CASE

Sequentially improving the clarity of materials.

23. Comparative / Multi-Option

EXAMPLE

“Give 3 ways to teach vocabulary (PPP, TBLT, Lexical) and list the pros/cons of each for an A2 class.”

USE CASE

Informing method selection by comparing pedagogical choices.

24. Contrastive Examples

EXAMPLE

“Provide one weak and one strong B1 speaking prompt; explain the key differences.”

USE CASE

Modeling and generating high-quality tasks by illustrating what to avoid.

25. Prompt Chaining

EXAMPLE

“Step 1: Analyze this test. Step 2: Identify the 3 most common errors. Step 3: Design two activities for each error.”

USE CASE

Creating a data-driven instructional sequence.

26. Clarifying Questions First

EXAMPLE

“Before creating the activities: What is the learner’s level, class size, L1, and available tech?”

USE CASE

Ensuring the final output is perfectly fit for the specific teaching context.

27. Critic/Reviewer + Revision

EXAMPLE

“Evaluate this worksheet for clarity and level; list any issues; then provide a revised version.”

USE CASE

Conducting a thorough quality assurance check on materials.

28. Scenario-Based Prompting

EXAMPLE

“You are a night-class coordinator. Design a 60-minute B1 speaking session for a class with mixed L1s and limited tech.”

USE CASE

Contextualized design for realistic and challenging teaching settings.

29. Localisation / Culture-Sensitive

EXAMPLE

“Adapt this ‘giving feedback’ role-play for a business English class in Japan, ensuring it is culturally appropriate.”

USE CASE

Region- and culture-aware adaptation of materials.

30. Retrieval-Augmented Generation

EXAMPLE

“Using the provided coursebook unit PDF, list the target grammar, vocabulary, and stated lesson aims for Unit 6.”

USE CASE

Grounding AI responses in specific documents for syllabus alignment.

31. Generated Knowledge Prompting

EXAMPLE

“First, list the key rules for using articles. Then, correct these 5 sentences and explain each correction based on the rules.”

USE CASE

Forcing the AI to build a knowledge base before applying it to a task.

32. ReAct / Tool-Use Prompting

EXAMPLE

“If a CEFR descriptor is needed, open the provided descriptors.pdf. If a word is unclear, consult the glossary. Then, finalize the lesson aim.”

USE CASE

Grounding lesson design in facts and specific data by allowing the AI to use "tools."

33. Tool/Function Calling

EXAMPLE

“If calculating the average score for this quiz, call the built-in calculator function, then present the result.”

USE CASE

Integrating external tools (calculators, search) within planning tasks.

34. Automatic Prompt Engineering

EXAMPLE

“Generate and test 5 different speaking prompts to elicit past narratives from A2 learners; select the most effective one.”

USE CASE

Automating the optimization of instructions and questions.

35. Self-Ask / Decomposition

EXAMPLE

“To design a lesson on this text: First, identify the main theme. Second, list 5 challenging vocabulary items. Third, propose a pre-reading question.”

USE CASE

Structuring the AI’s internal process for complex analysis or design tasks.

36. Meta-Prompting (System Prompts)

EXAMPLE

“You are an expert assistant for language educators. Prioritize accuracy, cite sources, and adopt a supportive, professional tone.”

USE CASE

Establishing global behavior and safety constraints for all interactions.

37. Anti-Prompting / Guardrails

EXAMPLE

“If a proposed activity risks an inclusivity issue, provide an alternative inclusive option instead of proceeding.”

USE CASE

Proactively enforcing safety, inclusivity, and policy constraints.

38. Memory/Stateful Prompts

EXAMPLE

“Remember: my learners are B1 adults, Spanish L1, and I prefer task-based lessons. Use this context for all future plans.”

USE CASE

Maintaining context and preferences across a conversation for personalization.

39. Non-Goals / Out-of-Scope First

EXAMPLE

“This lesson will not cover mixed conditionals; the focus is solely on spoken fluency with the second conditional.”

USE CASE

Preventing scope creep and focusing the AI's output.

40. Evidence-Tagging

EXAMPLE

“In your lesson rationale, mark research-backed points as [Cited: Nation, 2001] and tips from common practice as [Experience].”

USE CASE

Adding transparency to distinguish between evidence-based and anecdotal advice.

41. Multimodal Prompting

EXAMPLE

“Analyze this 30-second learner audio. Give 3 pronunciation priorities (IPA + minimal pairs) and 2 quick drills.”

USE CASE

Pronunciation feedback and practice using audio or other media.

42. L1 Contrastive Analysis

EXAMPLE

“For Spanish L1 B1 learners, list 5 common errors with articles/prepositions; give a targeted micro-task for each.”

USE CASE

Anticipating errors and tailoring practice by learners’ L1.

43. Backward Design Prompting

EXAMPLE

“Start from this B1 speaking rubric; derive lesson objectives, then select texts and tasks.”

USE CASE

Outcomes-aligned planning that begins with assessment.

44. Item Bank + Distractor Quality

EXAMPLE

“Create 10 B1 MCQs on travel vocab: 1 correct + 3 plausible distractors each; include distractor rationales.”

USE CASE

Producing higher-quality assessment items with analyzable distractors.

45. Branching Role-Play / Scenario

EXAMPLE

“Design a customer service role-play with 3 branches based on learner responses; include prompts and likely learner errors.”

USE CASE

Interactive speaking practice with contingencies.

46. Readability/Level-Controlled Text

EXAMPLE

“Write a 120-word B1 news summary: Flesch 70–80, no sentence > 18 words, 90% top-3000 words.”

USE CASE

Precise CEFR control and teachable lexis in texts.

47. Spaced Repetition Scheduling

EXAMPLE

“With these 20 B1 travel words, create a 2-week SRS plan (days 1/3/7/14) with micro-activities.”

USE CASE

Improving retention and recycling over time.

48. Back-Translation QA

EXAMPLE

“Write A2 task instructions; translate to Spanish; back-translate to English; revise until meaning drift <10%.”

USE CASE

Ensuring clear, multilingual-friendly instructions and catching ambiguity.

Crucial Distinction: AI Corrects, It Doesn't Evaluate

AI is excellent for generating rubrics, designing exit tickets, and correcting objective quizzes. However, evaluation is a deeply human process that requires empathy, understanding of a student's unique context, and pedagogical judgment. AI cannot replace the educator's role as the central evaluator and mediator.

49. Rubric Generation

EXAMPLE

“Act as an assessment expert. Create an analytical rubric for a B1 argumentative essay. Include 4 criteria (Task Achievement, Coherence, Lexical Resource, Grammar) across 4 performance bands.”

USE CASE

Standardizing evaluation criteria quickly and effectively.

50. Formative Exit Tickets

EXAMPLE

“Generate 3 creative 'Exit Ticket' questions to assess if my A2 students grasped the concept of countable vs. uncountable food nouns today.”

USE CASE

Quick, end-of-lesson formative assessment to gauge understanding.

51. Image Generation (Canva AI / Midjourney)

EXAMPLE

“A photorealistic image of a busy Brazilian open-air market (feira livre) with people buying fruits and vegetables, bright daylight, vibrant colors.”

USE CASE

Creating custom, culturally relevant visual flashcards for vocabulary.

52. Music Generation (Suno / Mureka)

EXAMPLE

“[Verse 1] I have been, you have gone, / We have traveled all night long. / [Style] Upbeat pop rock, clear vocals, educational.”

USE CASE

Creating custom, catchy grammar chants or vocabulary songs.

Advanced Techniques for ELT

Back-Translation QA

To ensure students don't misunderstand an AI-generated instruction, ask the model to translate the instruction into the students' L1 and then back into English. If the message comes back distorted, your original prompt needs clarity.

Few-Shot Prompting (Leveling)

AI doesn't "guess" what an A2 student is. Give examples.

"Here are 3 examples of typical A2 level sentences: [Ex 1], [Ex 2], [Ex 3]. Now, create a short text about 'holiday' following exactly this level of complexity."

Connect & Grow

EdTech ELT Community Follow me on Instagram Add me on Linkedin