Using AI to Create Texts for Classes

You open the coursebook. You read the text. And then you think, “I’m not using that!” And so begins the process of creating materials. There are usually two things you can do next. You might trawl the internet looking for a replacement. But it can be difficult to find an authentic text that matches what you’re trying to teach. If your search comes up blank, you only have one other option: write something yourself. This can be time consuming, and the results aren’t always great. But now there’s a third option. AI can write texts for us.

Free Online Workshop - Bring AI into Your Classroom: A Beginner's Guide for Language Teachers - Thursday, 9 May

In this blog, we’ll look at how you can use AI to create better texts faster. We’ll also explore some reasons for using texts and how these influence the texts we create. Finally, we’ll look at some free tools for checking that your AI-produced text is at the right level. First, let’s look at why it’s a good idea to use AI to create texts.

Why use AI to create texts?

Here are five reasons:

1. It’s fast. AI can create a useable text in a few seconds.
2. It’s surprisingly creative. AI can write a text using specific items of vocabulary and grammar structures. This would take most teachers (myself included) a long time.
3. It uses natural language. Because of the way AI is trained, the texts that AI creates sound natural. That means most texts created by AI will help expose learners to real-world language. If you’re worried about your own written English you can rest assured that any AI produced text will be error free.
4. AI can generate different kinds of texts. You can ask AI to write biographies, dialogues, diaries, monologues, fables, fairy tales, introductions to science, emails, letters, news reports, plays, poems, questionnaires, short stories, social media messages and lyrics to songs. As we’ll see later, these different text types lend themselves to practicing different reading skills.
5. An added bonus. Whatever you create will not be limited by any copyright laws.

Just to be clear, in this blog post I’m not saying that teachers should replace every text in every coursebook with texts made by AI. I’m simply saying that if you need a replacement text, AI can produce something of quality, quickly.

What are texts for?

Before we go any further, let’s pause for a moment to think about why we use texts in language classes. As we’ll see later, it’s important to know what you want to use a text for before you create one. Texts are so common in language teaching that we rarely stop to think about why we use them. So why do language teachers use and why do coursebooks contain so many texts?

The most obvious reason we use texts is to model language. Maybe you need to teach some new vocabulary. Why not present the new vocabulary in a story? By presenting language in a story, the students get exposed to vocabulary in context before you take out any flashcards or start drilling. But modelling vocabulary is not the only reason you might use a text in class. Some other common reasons are:

• To present grammar in context. For example, you might use a story which has several instances of the past continuous before going over the structure.
• To practice pronunciation. For example, you might read aloud a poem that has multiple examples of the sounds /e/ and /eɪ/ if your students have trouble distinguishing between these sounds.
• To show students an example of ‘good’ writing (sometimes called an “exemplar”). For example, you could show your class some simple, but well written restaurant reviews before asking them to write their own.
• To practice a specific reading skill or strategy. For example, you might want your students to scan a newspaper article to find some specific information.
• To practice writing skills or strategies. For example, you create an essay, delete all the discourse markers, then ask students to read and add their own. This would help students create cohesion when writing.

Next, let’s look at how to use AI to create texts for different purposes.

Creating texts using AI

Presenting language in context

Imagine that you have a list of vocabulary to teach and want to present this to your learners in context. We can ask AI to take the vocabulary list and use it to create a text. While the text might not be at exactly the right level, the creativity is priceless. To do this, you can use the following prompt:

Prompt: Write a [number] word article that uses the following vocabulary in a [topic] context. Use short simple sentences. Vocabulary: ### [your vocabulary].

Or prompt: Write a dialogue with [number] turns per speaker. The dialogue should use the following vocabulary in a [topic] context. Use short sentences and natural spoken language. Vocabulary: ### [your vocabulary].

Pronunciation practice

AI can create texts that help students practice pronunciation. First, create a text with several examples of a sound that learners have trouble pronouncing.

Prompt: Create a text that has several examples of these sounds. Sounds ### /[phonemic script]/ as in "[example word]" and /[phonemic script]/ as in "[example word]".

Then ask students to read these aloud using a speech-to-text AI (such as Otter). Students can compare the transcript produced by the speech-to-text AI to the original and aim to get as close to the original as possible.

Writing exemplars

Creating examples of writing for students to mimic is tricky. You don’t want to create something too complex. That would set the bar too high for your students. You also don’t want something too simple. That wouldn’t provide the class with enough of a challenge. This means you need to carefully craft your prompt to include all the necessary details. Some key things to include are:

• The word limit. This should be a similar length to what you expect your students to write.
• The genre. It is essential that the AI knows what style to write in. Is this a diary or a dialogue? A short story or a social media message?
• The complexity. What level of writing are you looking for?
• The language. Are there any specific words or phrases you want to include in the writing? What about grammar structures?
• What to avoid. Tell the AI what not to write. Perhaps complex grammar or a topic that would be culturally inappropriate for your students.

Putting it all together, we get something like the following:

Prompt: I want you to write a [genre] about [topic] using between [word limit range]. Use short simple sentences. Avoid [grammar, layout, topics, etc.].

After creating and refining a positive writing example, you can then ask learners to analyze the writing and notice features of the text. The students can then try to use some of these features in their own writing. Alternatively, give students a writing prompt and ask them to write a draft. After finishing the draft, students analyze the exemplar. Students note features that they can use to improve their drafts. Students improve their writing based on what they learned from analyzing the exemplar.

Reading skills practice

Teachers may want their students to practice reading for a specific reason. For example, you might be preparing students for a reading test. Or you may want students to practice a specific reading strategy, such as scanning. AI can help to create texts that match reading sub-skills. For example, imagine we need to create a text similar to one from an exam. We could ask the AI to analyze some examples and then create a new text.

Prompt: Please read and analyze this example text and note its style, lexical sophistication, voice, tone and register. Text: ### [Exam Text] ### Then write a new text about [topic] in the same style, voice, tone and register as the previous text.

Sometimes it is difficult to know what kind of texts match which kind of reading skills. For example, you wouldn’t normally skim music lyrics for gist, but you might skim an email. You might read an academic text and take notes, but you probably wouldn’t while reading a newspaper. AI can help by suggesting appropriate genres to match reading sub-skills.

Prompt: Tell me, which genre of texts do people commonly [reading subskill]?

From there, you can create texts for students to practice useful reading subskills such as skimming, scanning, previewing, annotating, making inferences and summarizing.

Writing skills practice

Writing is a complex process. Sometimes, you might want your students to practice a single area of writing in class. One way to do this is to give learners a text and ask them to do something with it. Below are some writing sub-skills and ways to practice these in class:  

• Cohesion. Ask AI to write a text. Remove all the pronouns and discourse markers. Ask students to add these back in. (Prompt: remove all pronouns and discourse markers from this text: ### [text]).
• Editing. Using AI, create a longer text than students would be required to write. Ask the students to edit this to make it shorter.
• Paragraphing. Ask AI to write an essay with the paragraphs removed. Get students to add the paragraphs.
• Punctuation. Ask AI to write a text. Then ask the AI to rewrite the text, removing all punctuation (commas, periods, capital letters, etc.). Ask students to add these back in. (Prompt: remove all punctuation (commas, periods, capital letters, etc.) from this text: ### [text]).
• Writing a conclusion. Get the AI to write an essay. Remove the conclusion. Ask students to read the essay and add a conclusion based on the rest of the essay. After writing, students can compare their conclusion to the original.

Adapting AI content

AI rarely produces exactly what you’re looking for on the first attempt. If you don’t get what you want, you can try some of the following prompts to get the AI to adapt or rewrite what it has written. Below are some common problems with AI produced texts and suggestions for how to solve them using AI.

• Too long. Often the AI produced texts are not at the right word count. Prompt: Please rewrite using [number range] words.
• Too complex. AI isn’t (at the moment at least) great at grading language. If you get a text that is too complex, try the following prompt. Prompt: Please simplify.
• Too many bullet points. AI loves bullet points. (So do I!) But these might not always be appropriate for the type of text you want to create for your students. Prompt: Please rewrite without using any bullet points.
• It just doesn’t work. Sometimes, you try and try again. But no matter what you do, AI doesn’t get there. If this is the case, try another AI model. If you always use ChatGPT, try Claude. If you love using Bard, try Perplexity. And if you’re not sure what is going to work, try giving the same prompt to several different AI models at the same time, and go with whatever is closest.
• In the wrong format. If you’ve created a dialogue for your students, you’ll want to turn this into audio. Luckily, you can use text-to-speech AI (like Natural Readers) to generate realistic speaking in different accents. The speaking speed can be changed to match the level of your learners. After generating this audio, edit the results using free audio software (like Audacity).

Evaluating your texts

After you’ve created a text using AI, you’ll need to check that it’s at the right level for your class. While AI text generators are great at lots of things, they’re not currently great at grading their language. That means that you might end up with a text which is too difficult for your students. But how can you check?

Below are some of my favourite tools for analysing the difficulty level of a text.

• Text Inspector. This program looks at all the vocabulary in a text and tells you how common (or uncommon) it is. You can use this to find any unusual words and either pre-teach these, or just replace them with a simpler equivalent.
• Oxford Text Checker. This program is similar to Text Inspector. But instead of saying how common each word is, the Oxford Text Checker gives each word a CEFR rating.
• Hemingway Editor. This fantastic website analyses the grammar in a text and suggests a native speaker reading level. If your text gets a score of Grade 12, that means that your students will need a native speaker reading level of grade 12 to be able to understand this. Hemingway Editor also highlights sentences which are particularly difficult to understand.

You don’t need to use a website to check the difficulty level of a text. Your own experience and intuition are also vitally important. Remember to read whatever AI has created before giving it to your students. When reading, look out for:

• Low-frequency words
• Metaphors and idioms
• Complex grammar
• Unfamiliar aspects of culture
• Lexical density.

Conclusions

AI text generators are an efficient way for teachers to create customized texts for their classes. AI can produce initial drafts of texts for nearly any purpose far faster than you writing them from scratch. This saves you time. With the right prompts, AI can allow you to produce high quality texts which meet your learners' needs. Remember to experiment with different AI models, analyse the outputs, and iterate until you have a text that will work for your students.