Estimated reading time: 7 minutes
AI prompts for ELL teachers turn a general chatbot into a differentiation engine, if you know what to write. This is a copy-paste library: prompts you can paste straight into ChatGPT, Claude, or Gemini to generate lessons, worksheets, speaking activities, writing tasks, and assessments matched to your students. Use the builder below to generate one in seconds, then learn the techniques to shape your own.
Table of contents
AI Prompts for ELL Teachers: Build Yours in 20 Seconds
The hardest part of using AI is writing the prompt. Skip that: pick your task, standard, proficiency level, grade, and topic, then open the ready prompt in ChatGPT, Claude, or Gemini.
The Anatomy of a Good ELL Prompt
Quick answer: a strong prompt for English Learners always includes five things: the grade, the proficiency level (WIDA, ELPAC, or TELPAS), the standard or objective, the scaffolds you want (sentence frames, word bank, visuals), and the output format. Add a speaking component and an answer key, and you get classroom-ready material on the first try. Weak prompts leave these out and produce generic, off-level results.
| Include | Example |
|---|---|
| Grade | “Grade 4” |
| Proficiency level | “WIDA Level 2” or “ELPAC Expanding” |
| Standard / objective | “aligned to WIDA speaking standards” |
| Scaffolds | “sentence frames, a word bank, visuals” |
| Output format | “as a worksheet with an answer key” |
How to Use the Builder
These AI prompts for ELL teachers all start the same way: use the builder above to pick your task, standard, proficiency level, grade, and topic, and it writes the prompt for you. Open it in ChatGPT with one click, or copy it into Claude or Gemini. Treat the first output as a draft and refine it with a single line: “make it one level lower,” “add more sentence frames,” or “shorten the reading.” Never paste student names or data into a public tool.
Prompts by Standard: WIDA, TELPAS, ELPAC
Name your standard so the AI aligns the output. Examples you can paste and adapt:
- WIDA: “Align this activity to the WIDA ELD Standards for Grade [grade], Level [level], speaking.”
- TELPAS (Texas): “Align this task to the Texas ELPS and TELPAS speaking descriptors for a [beginning/intermediate/advanced] Emergent Bilingual.”
- ELPAC (California): “Align this lesson to the CA ELD Standards and ELPAC [Emerging/Expanding/Bridging] level.”
- Dual language: “Give this activity in both English and Spanish for a two-way immersion class at Grade [grade].”
Align Your Prompts to WIDA, TELPAS, and ELPAC
The fastest way to get standards-aligned output is to tell the AI what the student’s level actually looks like. The builder at the top of this page does this automatically when you leave Add the proficiency-level descriptor checked, or you can paste a level descriptor into any prompt yourself. Below are short, plain-language summaries you can copy; for the exact official wording, use the source links.
| Standard | Level | What the writing looks like (plain-language summary) |
|---|---|---|
| WIDA | Level 2 (Emerging) | Short phrases and simple sentences; everyday vocabulary with some general academic words; frequent support. |
| Level 3 (Developing) | Expanded sentences and connected ideas; general and some specific academic vocabulary. | |
| Level 4 (Expanding) | A variety of sentence types and connected paragraphs; specific and some technical vocabulary. | |
| TELPAS | Intermediate | Simple sentences on familiar topics; frequent errors; limited academic vocabulary. |
| Advanced | Connected sentences and paragraphs with grade-appropriate content and support; errors rarely block meaning. | |
| ELPAC | Expanding | Increasingly varied English for many purposes; expanded sentences with moderate support. |
| Bridging | Broad, precise English approaching native-like fluency with light support. |
New York note: New York has joined the WIDA Consortium. The final NYSESLAT was given in spring 2026 and WIDA ACCESS begins in spring 2027, so New York teachers should now align to WIDA (select WIDA / ACCESS in the builder).
Copy the official descriptors from the source: the WIDA Can Do Descriptors (used by New York and 40-plus states), the TELPAS Proficiency Level Descriptors (Texas), the California ELD Standards, and New York’s move to WIDA.
Techniques That Level Up Every Prompt
Beyond the five details, a handful of techniques reliably improve what any AI returns for ELL work. The first is role prompting: begin with something like act as an expert ESL teacher designing for a WIDA Level 2 class, which orients the model toward the right expertise and register. The second is few-shot prompting: paste one example of a sentence frame or item in the exact style you want, and the AI will match it, which is far faster than describing the style in words. The third is asking the model to self-check: after it produces a worksheet, add a line such as review your own output and confirm every item is appropriate for WIDA Level 2, and simplify anything that is too hard. This catches level drift that otherwise slips through. The fourth is negative constraints: telling the AI what to avoid, such as do not use idioms or cultural references a newcomer would not know, prevents common problems before they appear.
Used together, these techniques turn a decent prompt into a dependable one. They also make iteration faster, because when an output misses, you can usually fix it with one targeted follow-up rather than rewriting from scratch. The goal is not a single perfect prompt but a short, confident exchange that lands on classroom-ready material in two or three turns.
Ready Prompts by Task Type
Here are starting points you can paste and adapt for common ELL tasks. Fill in the bracketed details, and add your scaffolds and output format.
- Vocabulary: “Create a [topic] vocabulary set for Grade [grade] at WIDA Level [level], with 8 words, kid-friendly definitions, an image idea for each, and a matching activity.”
- Reading comprehension: “Write a [length] passage on [topic] at WIDA Level [level], then add 5 questions ranging from literal to inferential, with sentence frames for the answers.”
- Speaking role-play: “Design a partner role-play about [situation] for Level [level], with roles, sentence starters, a word bank, and a short self-assessment checklist.”
- Formative check: “Create a 4-question exit ticket on [objective] for Level [level] that shows me who has met the language objective, with an answer key.”
- Family message: “Draft a short, warm message to families about [topic] at a plain reading level, then give a Spanish version for review by a fluent speaker.”
Save the ones that work for you into a personal library so you never start from a blank prompt again, and remember to keep identifiable student data out of any public tool.
Frequently Asked Questions
What are the best AI prompts for ELL teachers?
Prompts that specify grade, proficiency level, standard, scaffolds, and output format. Adding those five details is what turns a generic response into classroom-ready, on-level material.
Which AI tool is best for these prompts?
ChatGPT, Claude, and Gemini all handle these prompts well. Try the same prompt in two and keep the better output. For speaking practice the materials cannot provide, use a dedicated tool like Telo AI.
Is it safe to use these prompts?
Yes, if you review outputs for accuracy and bias and never enter student names, IEP details, or scores into a public AI tool. Follow your district’s AI policy.
What prompt techniques improve AI output for ELL?
Four help most: role prompting (act as an ESL teacher for a WIDA Level 2 class), few-shot examples (paste one item in the style you want), self-checking (ask the AI to confirm every item fits the level), and negative constraints (tell it to avoid idioms or references a newcomer would not know).
How do I stop AI from producing off-level material?
Add a self-check line after the task, such as review your output and simplify anything too hard for WIDA Level 2. Pair it with a negative constraint against idioms or unfamiliar cultural references. Then review the leveling yourself, since your judgment of your students is the final word.
How many tries should a good prompt take?
Aim for two or three turns, not one perfect prompt. Start with the five details, then fix any miss with a single targeted follow-up like make it one level lower or add sentence frames. A short, confident exchange usually lands on classroom-ready material.
Related Resources
Conclusion
Save this page. With these AI prompts for ELL teachers and the five details that make them work, you can generate differentiated, standards-aligned ELL materials in ChatGPT, Claude, or Gemini in minutes. Then give your students the speaking practice no prompt can deliver, and the materials you built will have somewhere to land.
Sources: WIDA Consortium; ACTFL; U.S. Department of Education, Student Privacy.
The Prompt Ends Where the Conversation Should Begin
A well-built prompt is a small marvel: name the grade, the level, the standard, and the scaffolds, and a differentiated worksheet or speaking activity appears in seconds. For an ELL teacher, that library of prompts turns hours of prep into minutes.
Every one of those outputs is still text, though. A prompt can produce a role-play script and a set of sentence frames, but the file cannot take a turn in the dialogue. An English Learner can hold a perfect speaking worksheet and still not speak, because a single teacher across a full class can spare each student only a moment of real conversation. What the library cannot generate is a partner to talk with. The limit is capacity, not the prompt.
So teachers pair the materials these prompts create with a tool that supplies the talk. Telo AI is one example, giving each student individualized conversational practice aligned to WIDA, TELPAS, ELPAC, and the day’s topic, in English plus Spanish and French for bilingual programs.
See how prompt-built materials lead into real speaking practice: https://mytelo.ai/how-telo-works/
The best prompt in the library can script a conversation but never hold one. The teachers who get the most from this page use it to build the materials, then make sure each student has somewhere to actually do the talking those materials call for.