How to Create ESL Quizzes With AI

A genuinely ESL-specific quiz question is a different design problem than a native-English grammar question, because it's testing a learner who is still building a separate, developing language system, not a native speaker making an occasional slip. AI can build that kind of item around three things: distractors that trace to a student's first-language grammar, a difficulty level calibrated to a WIDA or CEFR proficiency band instead of a grade number, and item language simple enough that a content-area quiz measures the subject, not English fluency.

Say you write a grade 6 science quiz item with a relative clause buried in it: "Which of the following best describes the process by which plants convert sunlight into usable energy?" A native-English-speaking student and a WIDA Developing-level student can know the same amount about photosynthesis and still land on different answers, because one of them is also decoding "the process by which" as a separate task.

That gap, between what a student knows and what a quiz question's language demands, is the real subject of this guide.

Quick Answer: A genuinely ESL-specific quiz item does three things a generic grammar quiz doesn't: it builds wrong answers around documented first-language transfer patterns (interlanguage theory, Selinker 1972), it calibrates sentence complexity to a WIDA or CEFR proficiency level rather than a grade number, and it separates language-proficiency testing from content-knowledge testing so a wrong answer on a science or math quiz reflects a content gap, not a language one.

This guide covers:

• Why testing an English learner is a second-language-acquisition (SLA) problem, not a literacy-instruction problem, and how that differs from English or ELA quiz design
• L1-transfer distractor design, grounded in interlanguage theory and contrastive analysis
• Calibrating item difficulty to WIDA's six levels and CEFR's six levels, not grade level alone
• The validity trap: when a content-area quiz accidentally tests English proficiency instead of subject knowledge
• A step-by-step AI workflow, a tools comparison, and the mistakes that show up most often

Why an ESL Quiz Is an SLA Problem, Not a Literacy Problem

An ESL quiz assesses a student's progress through second-language acquisition, a distinct research field from the native-literacy instruction an English or ELA quiz is built around. That's not a semantic distinction. It changes what a "good" wrong answer looks like, what "grade-appropriate" even means, and what a low score is allowed to mean.

ESL, EFL, EL, MLL: Which Term Actually Applies

"ESL" (English as a Second Language) technically describes a student learning English while living in an English-dominant environment, distinct from "EFL" (English as a Foreign Language), a student learning English in a country where it isn't the dominant language.

In U.S. federal law, the Every Student Succeeds Act uses "English Learner (EL)" as the operative term (ESSA, 2015), and WIDA's own more recent materials increasingly favor "multilingual learner (MLL)" alongside EL. This guide uses "ESL" because it's the term most teachers search for, but a class roster, IEP, or state report is more likely to say EL or MLL.

How This Differs From an English or ELA Quiz

An English quiz assumes a student is still refining a language they already command; an ESL quiz assumes a student is actively constructing that language, one rule at a time. The table below lays the three side by side.

A wrong answer on an English quiz usually means a student hasn't yet internalized a rule of a language they already speak. A wrong answer on an ESL quiz can mean several different things at once, an unlearned rule, an L1-transfer pattern, or a language-proficiency gap standing in front of content knowledge the student may already have. The next three sections build a quiz that can tell those apart.

L1-Transfer Distractors: Building Wrong Answers Around a Learner's Interlanguage

A well-built ESL distractor traces to a documented pattern in how a student's first language shapes their developing English, not a random wrong guess. Linguist Larry Selinker named this developing system an "interlanguage": a rule-governed language a learner constructs on the way to full proficiency, systematic rather than simply broken (Selinker, 1972). Many of its most predictable errors come from language transfer, patterns carried over from a student's L1 (Odlin, 1989).

Contrastive analysis, the practice of comparing a student's L1 and English structurally to predict likely errors, gives that transfer real, checkable shape (Lado, 1957). Two patterns generate the most testable, teacher-usable distractors: article use and word order.

Article Omission: A Distractor Set Built From L1 Grammar

Russian, Japanese, Korean, and Mandarin Chinese have no grammatical articles at all, which makes article omission one of the most heavily documented transfer errors among students from those language backgrounds. Take the item stem "___ dog is barking in the yard."

A weak prompt like "write an article-usage question" tends to return four options with no traceable logic. Naming the omission pattern directly, and which L1 backgrounds it's likely to show up for, is what keeps a generated item diagnostic instead of a lucky guess.

Word-Order Transfer: SOV and VSO Distractor Surfaces

English is a Subject-Verb-Object (SVO) language. Japanese, Korean, and Turkish are Subject-Object-Verb (SOV); Arabic and Classical Hebrew can be Verb-Subject-Object (VSO) (Odlin, 1989). A word-order item can put that structural difference to direct diagnostic use.

Take the target sentence "She reads a book every night" and ask a student to identify the correctly ordered version.

That single item does more diagnostic work than a generic "fix the sentence" prompt, because each wrong answer points to a specific, nameable structure rather than a vague sense that something is off.

A Proficiency-Calibrated Illustration

Say you teach a grade 4 class with several students whose home language is Korean. A two-option version of the article item, correct versus the no-article omission surface, tests the real pattern without adding the definite/indefinite and phonological distractors a WIDA Entering-level student hasn't built enough English to weigh yet.

A grade 8 class with mixed L1 backgrounds at a higher proficiency band is closer to ready for the full four-option word-order item, where SOV, VSO, and adjunct-placement distractors all carry genuine diagnostic weight rather than overwhelming a student who's still assembling basic sentence structure.

Calibrating Item Difficulty to WIDA and CEFR, Not Grade Level

A useful ESL quiz item is calibrated to a student's language-proficiency level, not their grade, because two students in the same grade can sit at entirely different points on the English-acquisition path. WIDA and CEFR are the two frameworks that give that calibration a real, checkable scale.

WIDA's Six Levels, and What Each Can Handle

WIDA's English Language Development Standards frame proficiency as six levels, Entering through Reaching, each describing what a student can typically do with language, not just how long they've studied it (WIDA, 2020).

CEFR's Six Levels, and Where It's Used Instead of WIDA

The Common European Framework of Reference for Languages (CEFR) describes proficiency across six bands, A1 through C2, and is the framework most commonly used outside the U.S. and in many international or dual-language programs (Council of Europe, 2001).

WIDA and CEFR were built independently and aren't officially cross-walked to each other; treat them as two separate scales rather than interchangeable labels. WIDA, tied to its ACCESS for ELLs test, is the framework in use across WIDA Consortium member states; CEFR shows up more often in international schools and non-U.S. contexts. Naming which one your school actually reports on is worth confirming before it goes into a prompt.

A Level-Calibrated Illustration

Say you teach a grade 7 class with a student newly arrived at WIDA Entering. A vocabulary-matching item with picture support, rather than a sentence-level grammar item, is the right calibration; a four-option L1-transfer item would test frustration tolerance more than English.

A grade 7 student at WIDA Bridging or CEFR B2 is a different case entirely. At that level, distractor design starts to resemble the native-English quiz item design covered in How to Create English Quizzes With AI, since the student's remaining gaps are closer to the ones a fluent peer might still be working through.

The Validity Trap: When a Content Quiz Tests English Instead of Knowledge

A content-area quiz written in English can measure a student's English proficiency instead of their subject knowledge, a well-documented validity problem in ESL assessment that's easy to miss because the quiz looks perfectly normal. A wrong answer on a science or math quiz doesn't automatically mean a content gap. It might mean the sentence carrying the content was the real obstacle.

BICS vs. CALP: Why Conversational Fluency Isn't Academic Fluency

Jim Cummins drew an influential distinction between two kinds of English proficiency: Basic Interpersonal Communicative Skills (BICS), the conversational fluency a student typically builds within a year or two, and Cognitive Academic Language Proficiency (CALP), the denser, more abstract language of academic content, which Cummins's research estimated can take five to seven years or longer to fully develop (Cummins, 1981).

A student can sound fluent at lunch and still be years away from the CALP a science or social studies quiz actually requires. That gap is invisible in casual conversation and easy for a teacher to miss without deliberately checking for it.

Linguistic Modification: Separating the Language Load From the Content

Researcher Jamal Abedi's work on testing English learners found that simplifying a test item's language, without touching its content difficulty, changes how English learners score (Abedi, 2004). In one study, Abedi and Lord found that English learners' scores rose on linguistically simplified versions of math word problems, while native-English-speaking students' scores on the same simplified items barely moved, evidence that the original items' language, not the math itself, was suppressing ELL performance (Abedi & Lord, 2001).

That targeted simplification has a name in the assessment literature: linguistic modification, rewriting a test item's sentence structure and vocabulary load while holding its content difficulty constant.

Building Content Quizzes That Isolate the Content

A few concrete habits keep a content-area quiz measuring the subject it claims to measure:

• Shorten sentences and avoid embedded clauses in the question stem itself, reserving complex syntax for items that are actually testing language.
• Add visual supports, a diagram, a labeled image, an icon, wherever a word alone is carrying meaning a picture could carry instead.
• Provide a word bank for the content-specific vocabulary the item requires, so a wrong answer isn't just a missed term.
• Separate language-proficiency items from content-knowledge items explicitly when both need checking, rather than folding them into one score.
• Ask a colleague or an EL specialist to flag any sentence a student could get wrong for reasons that have nothing to do with the subject.

A Step-by-Step AI Workflow for ESL Quizzes

Building an ESL quiz well starts with naming the proficiency level and the content-versus-language goal before writing a prompt, since a generic "ESL quiz" request defaults to grade-level language.

1. Name the WIDA or CEFR level directly, not just the grade, so sentence complexity has an actual target.
2. Decide whether the item is testing language or content. A grammar item can lean on L1-transfer distractors; a science or math item needs linguistic modification instead.
3. Name the student's L1 when it's shared by the student or a small group, so article, word-order, or vocabulary distractors can trace to a real transfer pattern rather than a generic guess.
4. For a content item, request linguistically modified language explicitly: short sentences, no embedded clauses, a word bank for subject-specific terms.
5. Request an answer key that names why each wrong option fails, an L1-transfer pattern, a proficiency mismatch, or a genuine content gap, so results are diagnostic.
6. Check every grammar call and every "linguistically modified" item by hand. A simplified sentence can accidentally change the content's meaning, not just its complexity, and that's worth a second read before it reaches a student.

You could use EduGenius to hold a class profile's WIDA or CEFR level and home-language notes while drafting a batch of L1-transfer and linguistically modified items from a single prompt, then export the set to PDF, DOCX, or Google Forms once every item checks out.

Tools for Building ESL Quizzes

A tool built for ESL quiz design needs to hold a proficiency level and a home language across prompts, and support linguistic modification for content items, three things a generic quiz generator doesn't ask for by default.

The wider assessment-format landscape sits in AI-Powered Assessment and Quiz Generation. For proficiency-tiered worksheet prompts, including how to name a WIDA level and an idiom exclusion in the request itself, see How to Generate ESL Worksheets With AI. For non-ESL, subject-specific quiz distractor design in a world-language classroom, see How to Create Spanish Quizzes With AI.