How Do You Teach a Mixed-Level ESL Class When You Are the Only Teacher? AI Differentiation Strategies

Fifteen students. Three complete beginners who arrived last month from countries where English instruction was minimal. Five upper-intermediate learners who watch Netflix in English and read young adult novels for fun. The remaining seven are scattered somewhere in between — some strong in speaking but weak in writing, others the reverse. No teaching assistant. Forty-five minutes. Go.

This is not a hypothetical worst case. This is Tuesday morning for a significant proportion of ESL teachers worldwide. Mixed-level classes are not an occasional challenge to be managed — they are the default operating condition of language teaching. And the gap between what pedagogical theory says you should do (differentiate everything) and what is physically possible when you are one person with limited preparation time has been, until recently, unbridgeable.

That gap is closing. Not because teachers have suddenly been given more time or smaller classes, but because AI tools have made it possible to generate differentiated materials at a speed that matches the reality of the job.

Why Mixed Levels Are the Norm, Not the Exception

If you have ever wondered why your class has such a wide proficiency range, the answer is usually structural rather than pedagogical. Mixed levels are created by the systems around us, not by any failure in teaching or placement.

Small school groupings. In many language schools — particularly smaller ones or those in less urban areas — there simply are not enough students to create homogeneous proficiency groups. A school with forty enrolled students cannot run six different levels. They run two or three, and each group contains a range.

Cost-driven tutoring groups. Private tutoring is expensive. Group lessons are more affordable. Parents and students form groups based on scheduling convenience, not proficiency alignment. The result: a tutor facing four students at four different levels, expected to serve all of them equally.

Online group classes. The economics of online teaching push toward larger groups. Platforms and schools fill classes based on availability and time zones rather than precise CEFR placement. A "B1" class may contain students ranging from strong A2 to weak B2.

Inadequate placement testing. Many language schools use placement tests that are poorly designed, outdated, or too blunt to distinguish between sub-levels. A student who tests as "intermediate" on a 20-question multiple-choice placement test might be anywhere from A2+ to B1+ in reality. And a placement test that measures grammar and vocabulary says nothing about speaking or listening proficiency.

Continuous enrolment. Schools that accept new students throughout the term inevitably create mixed levels within existing classes. The student who joins in Week 8 is at a different point from the student who started in Week 1, even if they tested at the same level on entry.

None of these factors are within the classroom teacher's control. But the teacher is the one who has to make it work.

The Traditional Approach (and Why It Fails)

Before looking at AI-powered solutions, it is worth being honest about why the traditional approaches to mixed-level teaching are unsatisfying.

Teaching to the middle. This is the path of least resistance, and most ESL teachers have done it at some point — calibrate the lesson to the average level in the room and hope for the best. The problem is obvious: the stronger students are bored and under-challenged, the weaker students are lost and anxious, and only the narrow band in the middle is actually learning at the right level. Over time, this approach widens the gap rather than closing it.

Creating multiple versions manually. This is what the textbooks recommend: prepare differentiated materials at two or three levels for every activity. In theory, it is excellent pedagogy. In practice, it means tripling your preparation time for every lesson. A teacher with twenty contact hours per week simply cannot produce sixty hours' worth of materials. The maths does not work.

Pairing stronger with weaker students. Peer teaching can be valuable, but using it as a primary differentiation strategy places an unfair burden on advanced students. They become unpaid teaching assistants, their own learning stalls, and resentment builds. It also assumes that a student who is more proficient in English is automatically a good teacher — a connection that does not reliably hold.

Giving advanced students "extra" work. Finishing early should not be rewarded with more of the same. Extension work needs to be genuinely extending — offering new challenges and deeper thinking, not just additional volume. Creating meaningful extension activities on top of the core lesson adds yet another preparation burden.

Strategy 1: Same Topic, Multiple Levels

The most powerful differentiation approach is also the simplest in concept: teach the same topic to the whole class, but provide materials at different CEFR levels so that every student engages with the content at their appropriate level of challenge.

AI makes this practical by generating parallel versions of the same material simultaneously. You provide one topic — say, "food and nutrition" — and the AI produces a reading passage, vocabulary activities, and discussion questions at A1, A2, B1, and B2. The core content is the same. The linguistic complexity varies.

This preserves something that many differentiation approaches sacrifice: shared learning. When every student is working with the same topic, whole-class discussions remain possible. The A1 student can contribute simple observations. The B2 student can offer analysis. Everyone is part of the same conversation, just at different levels of linguistic sophistication.

The practical workflow looks like this:

1. Choose your lesson topic and primary learning objective
2. Generate parallel materials at two to four CEFR levels
3. Print or distribute the appropriate version to each student or group
4. Conduct shared whole-class activities (introductions, discussions, presentations) where all levels participate
5. Assign level-appropriate individual or small-group work during practice phases

The key insight is that you are differentiating the language, not the thinking. A beginner can think critically about nutrition — they just need simpler language to access and express those thoughts.

Strategy 2: Tiered Worksheets from a Single Input

Worksheets remain a staple of ESL teaching, and for good reason — they provide structured practice, can be completed independently, and create a tangible record of learning. The problem is that a worksheet designed for one level is wrong for every other level in a mixed class.

AI solves this by generating tiered worksheets from a single prompt. You specify the topic, the skill focus, and the target levels, and the AI produces three versions:

Foundation tier. Heavily scaffolded: word banks provided, sentence starters given, matching exercises rather than open production, visual supports, simplified instructions. Designed for A1-A2 learners who need maximum support to engage with the content.

Standard tier. Moderate scaffolding: some vocabulary support, guided questions, a mix of recognition and production tasks. Designed for A2-B1 learners who can work semi-independently but benefit from structure.

Extended tier. Minimal scaffolding: open-ended questions, production-focused tasks, analysis and evaluation prompts, academic vocabulary. Designed for B1-B2 learners who are ready for genuine challenge.

All three tiers share the same learning objective, the same topic context, and the same core vocabulary — they differ in the level of support, the complexity of expected output, and the cognitive demand.

The time saving is substantial. Creating three worksheet versions manually takes most of an evening. Generating them with AI takes minutes. The teacher's role shifts from production to quality control — reviewing the output, adjusting for their specific students, and making the pedagogical decisions that AI cannot make.

Strategy 3: Flexible Assessment

Assessment in a mixed-level class is where differentiation gets hardest. How do you test students at different proficiency levels fairly? If everyone takes the same test, it is too easy for some and impossible for others. If everyone takes a different test, you cannot compare results or track programme-wide progress.

AI enables a middle path: assessments that test the same underlying competency at different CEFR levels. The construct being measured — reading comprehension, grammatical accuracy, vocabulary range, writing coherence — remains constant. The linguistic demands of the test items vary.

For example, a reading comprehension assessment on the topic of "climate change" might include:

• A1-A2 version: A simplified text (150 words) with high-frequency vocabulary, supported by images. Questions test literal comprehension: who, what, where, when. Response format: multiple choice and short answer with sentence starters.

• B1 version: A moderately complex text (350 words) with some academic vocabulary glossed. Questions test both literal and inferential comprehension. Response format: short answer and brief paragraph responses.

• B2 version: An authentic or near-authentic text (500+ words) with no vocabulary support. Questions test inference, analysis, and evaluation. Response format: extended written responses.

All three versions assess reading comprehension. All three produce valid data about whether the student is progressing within their current CEFR band. And because the assessments are generated by AI, the teacher does not spend three evenings creating them.

For a deeper look at how to design and generate differentiated assessments, see our AI assessment creation guide.

Strategy 4: AI Marking Across Levels

Creating differentiated materials is only half the challenge. You also have to mark them — and marking work at multiple proficiency levels against different criteria is one of the most time-consuming tasks in mixed-level teaching.

AI marking tools address this by applying different rubrics to different levels automatically. An A1 student's paragraph is assessed against A1 descriptors — basic vocabulary use, simple sentence structure, comprehensibility. A B2 student's essay is assessed against B2 descriptors — coherent argumentation, varied sentence structure, appropriate register, accurate complex grammar.

The marking is instant. The feedback is detailed and level-appropriate. And — critically — the teacher reviews everything before it reaches the student.

This human-in-the-loop approach is essential. AI marking is a first pass, not a final judgment. The teacher checks that the marks are fair, the feedback is accurate, and the assessment captures what the student was genuinely trying to communicate. Sometimes an A1 student produces a piece of writing that is grammatically rough but conceptually brilliant — the teacher catches that. The AI provides the efficiency; the teacher provides the discernment.

For ESL teachers who are marking across three or four proficiency levels in a single class, the time saving is transformative. What previously required an entire weekend of marking can be reviewed and finalised in an hour. That reclaimed time goes back into planning, professional development, or — just as importantly — rest.

For a detailed comparison of AI and manual marking approaches, see our guide on AI vs manual grading.

Strategy 5: Station Rotation with AI Materials

Station rotation is a classroom management approach where students move between different activity stations in small groups. Each station offers a different task, and students rotate on a timed schedule. It is particularly effective in mixed-level classes because each station can operate at a different proficiency level.

The barrier to station rotation has always been materials. Running four stations means preparing four different activities for every lesson — a preparation load that makes it impractical for daily use.

AI removes this barrier by generating all station materials from a single topic prompt. You design the station structure once, and the AI populates each station with level-appropriate content.

A typical station rotation for a mixed-level ESL class might look like:

Station 1 — Reading. Levelled reading passages on the lesson topic. Each group receives the text at their CEFR level. Comprehension questions are level-appropriate.

Station 2 — Writing. Writing prompts at different levels of scaffolding. A1-A2 students get sentence frames and word banks. B1-B2 students get open prompts with success criteria.

Station 3 — Speaking. Discussion cards with levelled questions. Simpler cards use closed questions and provide model answers. More advanced cards use open questions and require justification.

Station 4 — Grammar/Vocabulary. Targeted practice at each level. The grammar focus connects to the lesson topic, and the difficulty matches the group's proficiency.

The teacher circulates, spending focused time with each group. Because the materials are self-contained and level-appropriate, students can work productively at each station without constant teacher support — freeing you to give targeted attention where it is needed most.

Practical Example: One Lesson, Four Levels

Let us walk through a complete 50-minute lesson plan for a mixed-level class with students ranging from A1 to B2. The topic is "Travel and Transport."

Opening (10 minutes) — Whole Class

Show a set of transport images (bus, train, bicycle, airplane, boat). Elicit vocabulary from the whole class. Write key words on the board. Ask simple questions that every level can engage with: "How do you get to school?" "What is your favourite way to travel?" This activates prior knowledge and establishes the topic for all students.

Differentiated Practice (25 minutes) — Level Groups

A1 Group: Simple vocabulary matching — match transport words to pictures. Gap-fill exercises with a word bank: "I go to school by ___." Listen-and-point activity using an audio clip describing a journey. The focus is vocabulary acquisition and basic sentence structure.

A2 Group: Short paragraph reading about someone's daily commute (150 words). Simple writing task: "Describe your journey to school" using a provided sentence frame ("First I _, then I _, finally I _"). Basic past tense introduction: "Yesterday I went to _ by ___."

B1 Group: Article reading about transport in different cities around the world (350 words) with inference questions: "Why do you think more people cycle in Amsterdam than in Los Angeles?" Writing task: a blog post about a memorable trip (150 words, no scaffolding). Grammar focus: reported speech — "She said she had travelled by train."

B2 Group: Opinion article on sustainable transport (500 words) with critical analysis questions: "To what extent does the author's argument rely on economic rather than environmental evidence?" Essay preparation: outline an argument for or against banning cars from city centres. Debate preparation: prepare three arguments and anticipate counter-arguments.

Closing (15 minutes) — Whole Class

Bring the class together for a discussion using levelled question cards:

• A1 cards: "What is your favourite transport? Why?" (one-sentence answer expected)
• A2 cards: "Tell us about a journey you took. Where did you go? How did you travel?" (short narrative expected)
• B1 cards: "What do you think is the best way to travel in your city? Give reasons." (opinion with justification expected)
• B2 cards: "Should governments invest more in public transport or electric cars? Why?" (argument with evidence expected)

Collaborative poster: each group contributes one section to a class poster about "Transport in Our City" — drawings and labels from A1 students, short descriptions from A2, paragraphs from B1, and an opinion piece from B2.

Every student participates. Every student is challenged at their level. And the entire lesson's differentiated materials — four reading texts, four worksheet versions, discussion cards, and writing prompts — were generated from one topic input in under ten minutes.

When Resources Don't Match Your Context

There is a dimension to the mixed-level challenge that rarely gets discussed in mainstream teaching literature: context mismatch.

Most published ESL materials — textbooks, worksheets, online resources — are designed for Western classroom contexts. The topics assume familiarity with Western cultural references. The scenarios presuppose a certain kind of daily life. For teachers working in Southeast Asia, the Middle East, sub-Saharan Africa, or Latin America, these materials can feel irrelevant to students whose lives look nothing like the world depicted in the textbook.

This matters for mixed-level teaching because relevance is a significant factor in comprehension. A student who is unfamiliar with the cultural context of a reading passage faces a double barrier — linguistic complexity plus cultural unfamiliarity. In a mixed-level class, this compounds the existing proficiency gap.

AI addresses this by generating locally contextualised materials. Instead of a reading passage about taking the London Underground, you can generate one about riding a songthaew in Chiang Mai, a matatu in Nairobi, or a colectivo in Mexico City. The vocabulary is still English. The CEFR level is still calibrated. But the context connects to the students' lived experience, which improves both engagement and comprehension.

This contextualisation capability is particularly powerful when combined with the ESL teaching strategies that leverage bilingual support — you can generate materials that are both locally contextualised and bilingual, addressing two accessibility barriers simultaneously.

For teachers planning lessons with AI in international or multicultural contexts, the ability to customise cultural context is not a luxury — it is a fundamental requirement for effective teaching.