AI differentiation for multilingual learners: what works in the classroom

Teachers with multilingual learners in their classrooms face a persistent challenge that most instructional tools don't solve: the same content needs to reach students at genuinely different language proficiency levels, often in the same class period, without making anyone feel singled out.

AI differentiation tools can address this — not by replacing the teacher's instructional judgment, but by reducing the time it takes to create multiple versions of the same content. Here is what the research says, what works practically, and where AI tools fit.

What differentiated instruction means for multilingual learners

Differentiated instruction for English Language Learners (ELLs) and multilingual learners (MLLs) means adjusting how content is presented, how students engage with it, and how they demonstrate understanding — without changing the learning objective.

The ASCD framework for technology-supported differentiation identifies three dimensions of adjustment:

• •Content: What students are expected to engage with (text complexity, vocabulary load, visual support)
• •Process: How students make sense of content (scaffolded steps, partner work, translation support)
• •Product: How students show what they know (choice of modality, reduced linguistic demand for assessment)

AI tools are most useful in the content dimension — generating leveled versions of the same text quickly. They are less useful in the process and product dimensions, which require teacher knowledge of individual students.

The practical problem AI solves

The traditional approach to differentiating for MLLs involves rewriting texts, creating supplementary vocabulary lists, and building scaffolded question sets — for every reading, every lesson, every week. For teachers with four or five proficiency levels in a single class, that workload is not sustainable.

AI can generate a leveled version of any text in seconds. It can produce a simplified vocabulary list, create comprehension questions calibrated to different language demands, and translate explanatory text for students who are very new to English. What took 45 minutes now takes five.

Common Sense Education's 2023 review of technology for ELL students notes that tools with built-in differentiation features reduce the per-student planning burden for teachers while improving students' access to grade-level content — the goal of differentiation.

How the Wayground Chrome extension handles differentiation

Wayground's Chrome extension is particularly useful for in-the-moment differentiation of web-based content. A teacher or student can activate the sidebar on any web article, YouTube video, PDF, or Google Doc, and the extension generates leveled versions — easier, grade-level, and extended — plus comprehension questions calibrated to each level, flashcards for vocabulary, and an option to translate the simplified version into a home language.

For multilingual learners, this addresses the most common classroom scenario: the class is reading a shared text or watching a shared video, and some students need a scaffolded entry point while others are ready for the standard version. The sidebar handles the leveling without the teacher having to prepare three versions of the material in advance.

The accommodation features within Wayground's assessment platform extend this further. Teachers can assign an assessment and then set individual accommodations — extended time, text read-aloud, reduced choices, simplified language options — for specific students directly in the platform, without building a separate version of the assessment.

What AI differentiation cannot do

AI can adjust text complexity and generate comprehension questions at different language levels. It cannot:

• •Evaluate whether a student's silence reflects confusion, cultural communication norms, or language anxiety
• •Build the trust relationships that make multilingual learners feel safe taking risks in English
• •Understand the specific heritage language interference patterns that affect a particular student's English grammar
• •Make instructional decisions about when to push toward English-only and when to support native language use

These require teacher knowledge. AI differentiation tools work best when teachers use them to buy back time for the relational and observational work that only humans can do.

Research on technology-supported differentiation for ELLs

Three research threads are worth noting:

Shepherd's 2015 study, cited by 71 sources, found that technology that increases student access to differentiated content improves ELL achievement, particularly for students at intermediate proficiency levels. The effect size is moderate but consistent.

Cutter's research on technology use with ELLs found that tools that reduce language barriers without simplifying cognitive demand produce the best learning outcomes — students engage with grade-level thinking even when language scaffolding reduces the linguistic complexity of how they access content.

ASCD's 2023 review of differentiated learning and technology noted that the most effective technology-supported differentiation is teacher-directed: the teacher identifies who needs which scaffolds, then uses technology to generate them efficiently, rather than deploying undifferentiated technology tools to all students.

A practical classroom workflow

Here is a realistic differentiation workflow for a teacher with multilingual learners using AI tools:

1. Before class: Select the primary reading or video source. The night before, use the extension or another AI tool to generate a simplified version of the key text and a vocabulary list. This takes five minutes, not forty-five.

1. At the start of the lesson: Assign the grade-level version to most students. Assign the simplified version to students at beginning or early intermediate proficiency. Both groups work with the same content and same discussion questions — the language access point differs.

1. During independent or group work: Students who need translation support can use the sidebar to access home-language explanations of key concepts, while their peers work in English.

1. For assessment: Set individual accommodations in the platform before the assessment opens. Students with read-aloud accommodations receive the text read aloud; students with extended time have it automatically applied. The accommodations are private and do not require the teacher to manage them manually during the assessment.

1. After the assessment: Review results disaggregated by language proficiency level to identify patterns — not to lower expectations, but to identify where instruction needs to be adjusted for the next unit.

Can AI tools differentiate content for multilingual learners automatically?

AI tools can generate leveled text versions and vocabulary scaffolds automatically from any source material. They cannot assess individual student language proficiency or make instructional decisions about when scaffolding is appropriate. Teachers direct the differentiation; AI speeds up the material creation.

What accommodations are most effective for ELL students on assessments?

Research from Vanderbilt's IRIS Center identifies read-aloud, extended time, bilingual glossaries, and small-group testing settings as the most widely effective accommodations for ELL students on academic assessments. Digital platforms that administer these accommodations automatically reduce the logistical burden on teachers.

How do you differentiate a YouTube video for ELL students?

AI tools like the Wayground extension can generate comprehension questions at different language levels from any YouTube video, create vocabulary lists from key terms in the video, and produce a written summary in simplified English. Some tools also provide auto-generated captions that can be translated.

Is it appropriate to use AI to translate assignments for ELL students?

Translation can support access, but it is not a substitute for language development. For students who are very new to English, translation of key vocabulary and explanatory text supports comprehension of the content. Translating entire assignments prevents students from building English language skills. The appropriate use is targeted: translate to support access, not to replace engagement with English.

How can teachers track differentiation effectiveness for multilingual learners?

By monitoring assessment data disaggregated by language proficiency level over time. If ELL students' proficiency rates are growing on standards-aligned assessments while maintaining access to grade-level content, differentiation is working. Flat or declining rates suggest the scaffolding may be reducing cognitive demand rather than just linguistic demand.

What is the difference between modification and accommodation for ELL students?

An accommodation changes how a student accesses content without changing the learning objective (read-aloud, extended time, simplified language). A modification changes the learning objective itself (reduced content, different standards). Effective ELL differentiation uses accommodations, not modifications — the goal is access to grade-level learning, not a reduced version of it.

Common pitfalls to avoid with AI differentiation for multilingual learners

Over-simplifying content rather than language. Differentiation for multilingual learners should reduce linguistic complexity without reducing cognitive demand. An AI that simplifies a grade-level reading about climate change into a first-grade passage has removed the content alongside the language barrier. Students should engage with grade-level thinking even when the language scaffolding reduces the vocabulary load.

Using AI translation as a substitute for language instruction. Translation supports access. It does not develop English language proficiency. Students who receive all instruction and assessment in their home language do not progress through English language acquisition. AI translation is appropriate as a scaffold for students who are very new to English — not as a permanent workaround.

Not reviewing AI-generated simplified text before sharing. AI text levelers sometimes introduce errors when simplifying complex passages — changing a fact, misrepresenting a relationship, or losing a critical nuance. Teachers should review simplified versions before distributing them, especially for content-area texts where accuracy matters.

Treating all multilingual learners the same. "ELL" and "multilingual learner" cover an enormous range of proficiency levels, heritage languages, and educational backgrounds. A student who is a recently arrived newcomer has different needs than a long-term English learner who has been in US schools for six years. AI differentiation tools can generate multiple levels, but the teacher decides which level is appropriate for which student.

Documentation and compliance considerations

For students with IEPs or 504 plans that specify language accommodations, differentiated AI tools can support — but not replace — the formal accommodations required by law. A student whose IEP specifies "text read aloud in English" needs that accommodation provided consistently, regardless of whether an AI tool is also generating simplified text for the student.

Teachers should document which AI differentiation supports they are using for which students, particularly for students with formal plans. This documentation supports IEP and 504 compliance reviews and helps new teachers understand what has been effective for specific students.