😄 A student finishes a reading exercise and tells the teacher: "I understood every single word. I just didn't understand the sentence."
<summary placeholder>.
|
What this chapter covers
|
Reading materials are something many language teachers create regularly — texts for comprehension, texts as a vehicle for vocabulary and grammar, texts adapted to a learner's profession or interests. AI can generate reading texts from scratch, adapt existing ones to different levels, and help you build a complete reading lesson around the text.
This chapter is organized around how reading lessons are typically built. We start with prompts for producing the text itself, then walk through three stages most teachers are already familiar with: before reading, while reading, and after reading. The framework is useful here because AI helps differently at each stage.
How Do You Prompt AI to Generate Reading Texts?
AI can generate a reading text at any CEFR level on virtually any topic — tailored to a learner's profession, interests, or current course unit — in under a minute. As mentioned under how to write prompts, specificity helps.
Write a short reading text (150-200 words) about a first day at a new job for a B1-level adult ESL learner. Use past simple and present perfect (structures we are currently working on).
Adding the grammar structures you're working on turns a reading text into integrated practice. You can also specify the genre or text type:
Write a short news article (200-250 words) about a local community garden project. Target language level: B2. Use formal but accessible language typical of a local newspaper.
Or
Write an informal email (100-150 words) from one friend to another, describing a recent trip to Italy. Target: A2-level English learner. Use short sentences and common vocabulary.
The text type matters because AI adjusts tone and structure accordingly. An email reads differently from a news article, which reads differently from a short story. Exposing learners to a range of genres is also a small but useful pedagogical move: each text type has its own conventions, and learners benefit from seeing more than a steady diet of "short articles."
A useful instruction to add to any text-generation prompt is what the learner will do with the text, not just what topic it covers. "A B2 text learners will use to identify causes and effects in environmental policy" gives the AI more to work with than "a B2 text about the environment", and you've already done part of the next planning step.
Professional and interest-based texts
As with vocabulary, personalizing reading materials to a learner's profession or interests can make a real difference in engagement. Try something like the following.
My student is a B2-level nurse learning English. Write a reading text (200-250 words) about a typical shift handover in a hospital. Use vocabulary and situations she would encounter in her work.
Or
My student is a B1-level teenager who is interested in video games. Write a reading text (150-200 words) about the history of a popular game franchise. Keep the language accessible and engaging.
Adventure and continued stories
One use case we've heard about from teachers is generating continued adventure stories, where each episode picks up where the last one left off or use the same characters. Young learners in particular stay engaged because they want to know what happens next.
There's a pedagogical case for this beyond engagement. Research on extensive reading — reading lots of text at a comfortable level for general meaning — has long pointed to it as one of the most effective ways to build reading fluency, vocabulary, and motivation. The historical limitation has been supply: there's only so much level-appropriate, learner-relevant text available in the world. AI changes that. A teacher can now produce essentially unlimited level-appropriate reading material for a specific learner.
To get started you may start with the regular details of the learner and something about the level or things you want in the text. It may help adding details of characters and plots that you think are interesting or engaging. However, if you don't have ideas about details to support the AI, you can also just ask for suggestions.
I want to write a continued adventure story for a B1-level young learner (12 years old) who is learning English. We are currently practicing past simple and past continuous. Suggest me some engaging characters and captivating plots to chose from and create complete prompts I can use for generating start of the stories.
The you can create a new prompt with the suggestions you prefer. However, most chatbots should also understand if you just indicate your choice. The chatbots will typically guide you. When you have started the story, a key prompting tip is giving the AI enough context about the story so far so it will produce something that seems a continuation. To exemplify see the prompt below.
I am writing a continued adventure story for a B1-level young learner (12 years old) who is learning English. The story is about a girl named Lena who finds a magic map in her grandmother's attic. In the previous episode, Lena followed the map to an old lighthouse where she met a talking seagull named Captain Gull. The episode ended with Captain Gull saying he knows where the treasure is hidden, but they need to cross the Whispering Forest first. Write the next episode (250-300 words). End with a cliffhanger. Use past simple and past continuous. Include 5-6 vocabulary words that would be new for a B1 learner and underline them.
As described in principles about prompting you should iterate if you're not happy with the result. Describe what is wrong or what you would like the AI to try to improve until you hopefully have something you're pleased with. If you do this regularly, it helps to keep a summary of the story so far that you paste into each prompt. Otherwise the AI may forget details or contradict earlier episodes. You may be able to get AI to make the summary and keep track of important parts by using a prompt like the following after the AI has given you the story you needed.
Summarize and include important parts and construct a prompt I can use to generate the next part of the story. Allow me to add a topic or indication of what I want from the text, but also keep it open so I can leave it out and have you come up with something relevant.
Many AI chatbots have the concepts of stored chats. It may be possible to just continue the chat where you have produced earlier episodes in your story. However, if you still experience issues, it may be a good idea to have the explicit summary. And getting AI to generate your next prompt may work. If you wish to focus on some specific vocabulary or grammar concepts later, you just need to include that in the prompt. Even if you started out with instructions to focus on some grammar, it is fine to switch later.
Illustrations
You can also pair episodes with illustrations to keep the story visually engaging. AI image generators struggle with character consistency across separate prompts, but repeating key visual details each time helps.
Generate an illustration for episode 3 of a children's adventure story. The main character is Lena: a 12-year-old girl with short brown hair, a red backpack, and green rain boots. In this scene, Lena and a large white seagull are standing at the edge of a dark forest. Style: children's book illustration. No text in the image.
The image generators are getting better and better, also at generating text in images. However, it may still sometimes look off. So if you don't need it, it may be a good idea to tell the AI. Even with detailed descriptions, AI may not produce perfectly consistent characters across images. Including the same key details each time helps, but you may need to regenerate to get something close enough.
How Do You Use AI to Adapt Existing Texts?
AI can adapt existing texts in three directions: simplifying for lower levels, elevating for higher levels, and repurposing the same content for a different skill focus or context. Rather than generating from scratch, you may want to adapt a text you already have, e.g. an article you found online, a passage from a coursebook, or something your learner brought from work.
Simplifying
If you need to simplify an existing text to match something your learner will not struggle too much with, you may use a prompt like the following. Depending on your learner and level, it may be helpful to provide some additional tips beyond just the CEFR level. You can always try with just the level, but then add additional instructions if it doesn't turn out like you wanted it.
Simplify the text below to A2 level. Keep the main ideas but use shorter sentences, simpler vocabulary, and remove complex structures. The text is for an adult ESL learner.
---
[paste the text]
AI handles simplification reasonably well, but watch for two failure modes. It may strip out nuance that mattered to the original meaning, and it may keep the same information density (just with simpler words), which doesn't actually make the text easier. True simplification reduces both linguistic and informational load. Try it out and iterate with specific instructions that address what it is doing wrong, if the adapted text is not satisfactory.
Making more complex
Simplification is the common direction in practice, but if you have a text that is a bit too simple for your learner, but want to use it anyway, you can try something like the following to make the text harder, and at your learner's level.
Rewrite the text below at C1 level. Add more sophisticated vocabulary, longer sentence structures, and more nuanced expressions while keeping the same content and meaning.
---
[paste the text]
Complexifying is harder than simplifying with AI. AI is generally (too) good at adding fluff and making texts longer, which may feel forced and not very interesting. Other times it may add content or change meaning, which may not be appropriate for instance if the text is not fictional. Again, the advice is to iterate and add specific instructions to address what you want.
Adapting for a specific purpose
If you have a good text, but want it in a different format, e.g. turning an article into a dialog or similar, you may try something like the following prompt.
Rewrite the text below as a dialog between two colleagues discussing the same topic. Keep the key information but make it conversational. Target level: B1.
---
[paste the text]
Other examples may be to turn aformal text into an informal one, a long text into a short summary, or rewriting something in another genre to make it more engaging.
One text, different tasks
A useful affordance of AI for differentiation: once you have a text that works, you can ask AI to produce different task sets for the same text at different levels. This helps for mixed-level groups, or for an independent tutor running back-to-back lessons with learners at A2 and B1 — same text, two task sheets.
For the text below, create two task sheets on the same content: - An A2 version with true/false questions and a vocabulary matching exercise - A B1 version with inference questions and a short summary task
---
[paste the text]
How Do You Use AI to Prepare Learners Before Reading?
What you do with a text before the learner sees it often matters more than the text itself. Pre-reading activities activate what the learner already knows about the topic, give them a reason to read, and front-load the vocabulary or context they need to make sense of the text. AI can generate pre-reading materials quickly, and they're some of the easiest wins in this whole chapter.
Activating prior knowledge
I am about to give a B1 learner a reading text about renewable energy sources. Generate 5 questions I can ask before she reads to activate what she already knows about the topic. Mix factual recall with personal opinion questions.
These can be asked in conversation at the start of the lesson, or written down for the learner to consider on their own. The point is to get the learner thinking about the topic in their own words before they meet the new vocabulary.
Predicting from clues
For the text below, generate a prediction activity: give the learner the title and 5 key words from the text, and ask them to guess what the text will be about. Include 3 follow-up questions they should try to answer from their prediction, which they can then check against the actual text.
---
[paste the text]
Predictions are useful even when they turn out to be wrong. Checking a wrong prediction against the text can be a strong comprehension exercise.
Pre-teaching key vocabulary
Dealing with unknown words may be part of reading, but if there are too many new words some learners may give up. Some words the learner may be able to guess or infer the meaning of, but some may be best to pre-teach.
From the text below, identify the 6-8 words most likely to block comprehension for a B1 learner. For each, provide a short learner-friendly definition and an example sentence. Format as a pre-reading vocabulary sheet the learner can review before reading.
---
[paste the text]
How Do You Prompt AI to Build Reading Comprehension?
Reading may include various skills and modes like skimming for the overall point, scanning for a specific detail, reading carefully when accuracy matters, and reading between the lines to catch what isn't said directly. Different question types provided up front my prime for and train different sub-skills, and AI can generate each on demand.
Skimming for gist
Skimming is fast reading for the overall message. Question types that train it are short and high-level: best title, one-sentence summary, main idea.
For the text below, create a skimming task:
- 3 possible titles, of which one is the best fit
- 2 one-sentence summary options, of which one accurately captures the text and one is plausible but wrong
- A brief instruction: "You have 60 seconds. Read quickly and pick the best title and summary."
---
[paste the text]
Scanning for specific information
Scanning is fast reading to locate a specific piece of information. Good scanning questions ask for concrete details: a number, a name, a date, a place.
For the text below, create 6 scanning questions. Each should ask the learner to find one specific piece of information: a number, name, date, place, or other concrete detail. The questions should not require understanding the surrounding context, just locating the information.
---
[paste the text]
Reading for detail
Detail questions are the workhorse: comprehension checks that the learner has actually understood what's there.
Create comprehension questions for the text below. Include:
- 3 true/false/not stated questions
- 3 multiple-choice questions
- 2 open-ended questions that require the learner to explain or infer
Target level: B1.
Include an answer key.
---
[paste the text]
Inferring
Inference questions ask the learner to read between the lines. The answer is supported by the text but not stated directly.
Create 5 inference questions for the text below.
- Each question should require the learner to read between the lines.
- The answer should not be stated directly in the text.
- Include an answer key with the relevant text passage that supports each answer.
---
[paste the text]
Inference and open-ended questions need more review than true/false or multiple choice, since the "correct" answer can be debatable. Having the AI include the supporting text passage in the answer key makes it easier to verify.
Recognizing text structure
How a text is organized is part of its meaning. Recognizing cause-and-effect, problem-and-solution, chronological sequence, comparison, helps learners understand what the text is doing, not just what it says.
For the text below, create a text-structure exercise:
- Identify the overall organization of the text (e.g., problem-solution, cause-effect, sequence, comparison)
- 4 questions that help the learner trace this structure (e.g., "what is the problem the author describes?", "what solution is suggested?")
- An answer key.
---
[paste the text]
Guessing vocabulary from context
Strong readers guess at unfamiliar words from the surrounding context rather than reaching for a dictionary every time. This is a reading strategy worth teaching explicitly. AI can build the exercise that walks the learner through it.
From the text below, choose 5 words a B1 learner is unlikely to know but where the surrounding sentences give enough clues to guess the meaning. Create an exercise that:
- Lists each word with the sentence it appears in
- Asks the learner what they think the word means and what clues in the sentence helped them
- Provides a learner-friendly definition in the answer key
---
[paste the text]
Some AI chatbots offer a feature often called "canvas" or similar, which can generate interactive web pages you can share with a link. Comprehension exercises work well in this format. The learner reads the text and answers on screen, with immediate feedback on true/false and multiple choice items. You can ask for this by choosing the mode or adding something like "Create this as an interactive web page where the learner can click answers and see their score" to your prompt.
How Do You Use AI to Extend a Reading Text After Class?
A reading task is only the start of what you can do with a text. The text becomes a launchpad for further work, personal response, re-reading for a new purpose, comparing perspectives, or feeding into other skills.
Personal and critical response
Reading shouldn't end at comprehension. Asking the learner what they think about the text, whether they agreed, what surprised them, what they would have done differently, turns reading into thinking and creates a natural bridge into speaking or writing.
For the text below, generate 5 personal-response questions a B1 learner can answer after reading.
- Each should invite the learner's own opinion or experience, not just "did you understand the text?"
- Include prompts that work for short verbal answers in conversation.
---
[paste the text]
Re-reading with a new lens
Re-reading a familiar text for a new purpose is one of the higher-value exercises in reading instruction. Comprehension deepens, and the learner notices things they missed the first time. AI can generate a second pass on the same text.
The learner has already read the text below for general comprehension. Create a second-pass exercise that asks the learner to re-read with a specific focus: the author's attitude toward the topic. Generate 4 questions that draw attention to word choice, tone, and what the author emphasizes or downplays.
---
[paste the text]
This is also a useful way to get more out of one text across multiple lessons.
Comparing texts
Reading two short texts on the same topic and comparing them builds critical reading, recognizing different perspectives, identifying bias, synthesizing information.
Generate two short texts (150 words each) at B2 level about [topic].
The two texts should take noticeably different angles (for example, optimistic vs. cautious).
After the texts, create 4 comparison questions asking the learner to identify where the texts agree, disagree, and emphasize different points.
Using the text as input for other skills
A good reading text is rarely just a reading text. The same text can feed vocabulary work, grammar mining, a writing follow-up, or a speaking task. The prompts for those activities live in their own chapters. What's worth flagging here is that once you have a text that works, it's worth holding onto and reusing across directions.
How Can AI Use Learners' Native Language for Reading Tasks?
For lower-level learners, L1 support can make reading texts more accessible. You can generate vocabulary lists and pre-teach the most challenging words, see the chapter on vocabulary prompts. You can also ask for a brief L1 introduction to the topic before the learner reads in the target language. This is pre-reading work done in the learner's L1. It may be useful for beginning learners who can't yet activate prior knowledge by talking about the topic in the target language.
My student is a Japanese-speaking A2 learner. Before she reads the English text below about climate change, write a 3-sentence introduction to the topic in Japanese. Keep it simple, the goal is to give her context so she can approach the English text with some background knowledge.
---
[paste the text]
The caveats from Chapter 3 apply. If you don't speak the L1, be transparent with the learner and let them verify.
What Should You Watch Out for When Using AI for Reading?
CEFR levels are approximate. AI tends to overshoot on reading texts — you ask for B1 and get something closer to B2. Specifying word count, sentence length, and vocabulary constraints helps, but always review the text yourself.
Specify word count. Without a word count, AI may produce a text that's too long or too short for your purpose. Being explicit ("150-200 words") keeps the output manageable. But be aware it only helps, it does not match your given word count exactly.
Check facts in generated texts. AI-generated reading texts sometimes include plausible-sounding but incorrect facts, especially about historical events, statistics, or geography. If factual accuracy matters, verify the key claims.
Adaptation changes meaning. When simplifying or complexifying texts, AI sometimes changes the meaning in subtle ways — removing a nuance, adding an implication that wasn't there, or dropping an important detail. Comparing the original and adapted version side by side helps catch this.
Set a purpose before reading. Reading without a purpose tends to drift. Even a quick "read this and find out X" framing improves engagement and comprehension. Build the purpose into your prompts, your pre-reading questions, or both.
Keep a story bible for continued stories. If you use the adventure story approach, maintain a summary of characters, plot points, and where the story left off. Paste it into each prompt — otherwise the AI may contradict earlier episodes.