Struggling to find decodable texts that match your classroom activities? Texts written for beginning readers are an important part of early literacy instruction. Decodable texts and alliterative texts help children learn phonics patterns (e.g., consonant-vowel-consonant words like “cap” or the /b/ sound at the beginning of words). Using AI, teachers can quickly generate texts aligned to what they’re teaching and to students’ interests, rather than relying on commercially available materials alone.
Once you’re familiar with the process, creating a custom text takes about 10 minutes—5 minutes working with the AI and 5 minutes for your own edits.
Many AI tools have free versions for teachers: We’ve had success with ChatGPT, SchoolAI, Gemini for Education, Claude, and MagicSchool AI. AI-generated decodable texts can be used within the typical array of instructional contexts: small-group, individual tutoring, and whole-group shared reading.
To use AI effectively for early readers, teachers need to understand both what makes a quality text and how to prompt and refine AI output.
Characteristics of Engaging Texts for Early Readers
Not all texts for early readers work equally well. Some are too dull despite focusing on target patterns; others include words or structures beyond young readers’ skills. Teachers should aim for a “sweet spot” where the text:
- Relates to children’s experiences and knowledge. For example, after students pop popcorn and observe how it jumps in the pot, a teacher might share this poem that reinforces the /op/ sound:
Hop to pop pop-corn.Drop pop-corn in a pot.Put a top on top of the pot.See pop-corn hop and pop.Do not let pop-corn hop out.Put a lid on top and hop back.
- Focuses intensively on the specific phonic targets or skills (e.g., short-a words or the /k/ sound).
- Is concise and focused. AI tools can default to wordy or rambling responses: “Once, I got a great present. My mother gave me a new cat! But the cat was so heavy, I could hardly pick it up. It was fat!” A more decodable version might be: “My mom got me a fat cat!”
- Includes high-frequency words the children are learning (e.g., put, get, got, is, the, I).
- May include some words beyond the phonic goals for the lesson. For instance, in a text focusing on consonant-vowel-consonant words related to a squirrel, children can read words like nut, get, and nap, while the teacher could read the word squirrel.
- Tells a coherent story or follows some other meaningful or predictable organization.
Strategies for Writing Prompts
Through our adventures using AI to write texts for a university-supported online early literacy project, we pulled together some tips for using AI to compose texts for early readers. These texts can take many forms: instructions, descriptions, stories about animals, stories with the reader as the actor, poems, chants, predictable texts, call-and-response games, or alliterative texts.
See Figure 1 for an example of our iterative prompting process. Don’t worry if this seems complex—the process becomes intuitive quickly.

Start with the Right Tools and Mindset
- Train an AI “assistant.” Most versions of AI programs offer this feature, which remembers previous conversations (e.g., Gemini Gems, Microsoft agents, SchoolAI’s “Dot”).
- Tell the AI to remember key information, such as previously successful prompts or information about your target audience. Keep in mind that there may be a limit on the number of facts an AI can recall, so input reminders as necessary over long chats.
- Drop in several high-quality texts and tell the AI to model its response on those examples. If you are not the creator of the example texts, make sure to secure permission from the author or publisher to upload them.
- Have a specific vision before you start: “Write a story about a spider with spots” works better than just “a text with sp words.”
Give AI Clear Parameters
Think of these as ground rules that guide the AI. The more specific you are up front, the better your results will be. Consider the following:
- State tense and/or helping verbs (e.g., can, will).
- Specify high frequency words that could be used for each grade level (e.g., is, get, put, the).
- Suggest key words that signal text structure (e.g., for compare/contrast text: and, but, is, not, has, does not have).
- Tell it to use content that is true in real life (e.g., things a dog can do like run and play, but not shout).
- Be specific regarding type of text; audience; tone of voice; type of materials; related activities; target words; grammatical elements to use or avoid; the actor in the text (e.g., first-person I, third person we, implied subject of commands); and stance (e.g., “Speak as if I’m a child”).
- Include “guardrails”: instructions about what you don’t want it to do (e.g., no words with final blends).
- State generally how many words in the text can vary from your target pattern (e.g., no words that don’t conform, hardly any words, a few words).
Iterate to Fine-Tune
- Treat AI as a brainstorming tool. The interaction should resemble a conversation.
- Keep trying, even if the first response isn’t great. Use longer, more detailed prompts for higher-quality responses.
- Check for mistakes—AI can make errors even with clear parameters (e.g., in one activity, an AI incorrectly identified “sled” as not containing a short-e sound).
- Tell it what you liked and didn’t like about the response.
- Ask, “How would you improve this prompt?” then copy-paste the response into the prompt field.
- Point out spots where the result is random or doesn’t make sense. Ask the AI to logically link the sections that are random.
- Note which prompts work, and ask your AI assistant to remember them.
Keep in mind, the first result is rarely perfect—that’s expected.
Finish by Hand
Once AI gives you a solid draft, finish it yourself rather than endlessly refining through prompts. For example, the MagicSchool AI platform recommends using the “80/20 Rule.” That is, let AI handle the initial 80 percent of your draft, then add your final touch as the last 20 percent.
From Practice to Routine
While creating AI-generated decodable texts requires some initial trial and error, teachers report that the process becomes faster with practice. The resulting texts, which connect directly to students’ classroom experiences and interests, can significantly boost engagement and make early reading practice feel more meaningful to young learners.







