Five point three million students in American public schools are learning English at the same time as they are learning everything else. That is the most recent count from the National Center for Education Statistics: 10.6% of all public school students, one in ten children in your building. Yet most AI-in-education coverage treats English language learners as an afterthought: a translation button, a checkbox on a pricing page. This post is for classroom teachers, ESL specialists and the district leaders who buy for them. If you teach multilingual learners (EAL learners, for UK readers), this is the shortlist I would hand a colleague.
One rule before the list. If the AI does the English, the student does not learn the English. A good tool lowers the barrier to practice; a bad one removes the need for it. Translation gets a student through the door of today's lesson. It does not teach them to open tomorrow's. Hold onto that rule and half the market filters itself out.
How I chose the tools
I started from the 90 tools listed on aieducator.tools, where 25 surface some form of English learner support, most clustered in the special education and accessibility and tutoring and personalised learning categories. I judged them on the jobs ELL teachers describe: levelled input students can read, structured speaking and listening practice, and materials families can access in their home language. I also weighed evidence of classroom use, safety posture, honest pricing, and whether the tool keeps the language practice with the student. No vendor pays to appear in the directory or in this post, and nobody can buy a placement.
What I left out matters as much as what made it. Duolingo is the name teachers search for most, and I excluded it: it is a consumer habit app, and a streak on a phone will not help a newcomer follow your science lesson. I also left out Wayground and SayVeritas, both tools I rate for other jobs, because translation support inside a quiz platform and formal oral assessment sit around ELL teaching rather than at the centre of it. If your question is broader differentiation rather than language support, I covered that in my guide to AI differentiation tools. Pricing and features verified 24 August 2026.
One number to hold before the table: across the whole directory, only 29 of 90 tools (32%) declare COPPA compliance, while 90% are usable by students under 13 on their stated minimum age. That gap is the first thing US schools should check before rolling anything out to young learners, and I unpack the whole compliance picture in Which AI Tools Are Safe for Schools?
The Comparison at a Glance
| Tool | Best for | Pricing | Min age | Privacy (declared) | Data hosted | Independent security assessment |
|---|---|---|---|---|---|---|
| Diffit | Levelled reading materials | Freemium | Not stated (teacher tool) | FERPA, COPPA (vendor site) | Not stated | No |
| MagicSchool | All-in-one district platform | Freemium | Not stated | FERPA, COPPA, SOC 2 (vendor site) | US (vendor site) | Yes (9ine) |
| Snorkl | Speaking practice | Freemium | Not stated | FERPA, COPPA (vendor site) | Not stated | Yes (9ine) |
| Kami | Accessible documents | Freemium | No minimum | FERPA, COPPA, GDPR | US | Yes (9ine) |
| Glint | Newcomer visual supports | Freemium | 18+ (teacher tool) | FERPA, COPPA, GDPR | US | No |
| EchoGuide Pro | Reading fluency | Paid | 4+ | FERPA, COPPA, GDPR | UK | No |
| Dictation Trainer | Listening practice (13+) | Freemium | 13+ | GDPR in progress | Global (Cloudflare) | No |
The privacy column is the vendor's declaration, not an audit; the final column is the independent one, drawn from 9ine's security assessments. "Not stated" means our directory record is blank, not that a tool is non-compliant; where a record was blank I checked the vendor's own privacy pages and say so below.
Best for levelled reading materials: Diffit
Diffit is the fastest way I know to turn any text your class must read into versions every student can read. Paste a text, a URL or a YouTube link and it produces the same content at multiple reading levels, with vocabulary lists, comprehension questions and translation into 70+ languages. That is the core ELL job: comprehensible input without watering down the content. What sets it apart from MagicSchool's text leveller is depth on this single workflow; scaffolds, graphic organisers and English learner supports are built around the levelled text rather than bolted on. Watch out for the free tier's limits: you can only reach content made in the last 90 days, and exports to Google Docs, Slides and Forms sit behind the school plan. Our directory record holds no compliance data for Diffit, so I checked: the vendor's own pricing and privacy pages state it is FERPA and COPPA compliant and collects no student data. Best for planning input across subjects; not for student practice. Pricing: free tier for teachers; Diffit for Schools is a flat-rate annual licence tiered by enrolment, quote-based, with individual teacher licences also available (verified 24 August 2026).
Best all-rounder with district controls: MagicSchool
MagicSchool is the safest bet if you want one platform covering planning, levelling and supervised student practice. Its 80+ teacher tools include a text leveller, translated family communication and IEP support, and its student rooms let learners use AI inside boundaries you set. It is one of only ten tools in the directory holding an independent 9ine security assessment, and the vendor documents FERPA, COPPA and SOC 2 compliance with US data hosting. What sets it apart from Diffit is breadth plus governed student access; the trade-off runs the other way too, because a platform with 80 tools does nothing as deeply as a specialist. Watch out for choice overload: without a plan naming the three tools your staff will use, it becomes a novelty by October. Best for schools that want one approved platform; not for a teacher who only needs levelled texts. Pricing: free plan for teachers; Plus at $8.33 per month billed annually, or $12.99 billed monthly; Enterprise quote-based (verified 24 August 2026).
Best for speaking practice and reasoning: Snorkl
Snorkl gets English learners talking, which is the practice most classrooms give them least. Students record their voice and draw on a whiteboard to explain their thinking, and the AI returns instant feedback in more than 40 languages against criteria you set. For an ELL student that changes something fundamental: they can show they understand the maths while their English catches up, then re-record in English when they are ready. It is used in 10,000+ schools, holds a 9ine assessment, and its own site documents COPPA and FERPA approval with a DPA and SSO on every tier. What sets it apart from SayVeritas is that Snorkl is a low-stakes practice space rather than a formal oral assessment platform. Watch out for the free tier's activity cap, and for feedback quality leaning heavily on the criteria you write. Best for oral language and reasoning; not for reading instruction. Pricing: free for teachers with limited activities; school and district plans quote-based (verified 24 August 2026).
Best for making every document accessible: Kami
Kami turns the PDFs and worksheets you already use into documents an English learner can work with: read-aloud, translation, a multilingual dictionary, voice typing and dyslexia-friendly fonts sit on top of any file. It declares FERPA, COPPA and GDPR compliance, hosts data in the US, holds a 9ine assessment and sets no minimum age, one of the cleanest safety profiles on this list. What sets it apart from Diffit is that Kami adapts the document in the student's hands rather than producing a new version of the text; the two pair well. Watch out for the fact that it is an annotation platform first: the language tools are the reason it is on this list, but you are buying the whole workspace, and the best of it sits in paid tiers. Best for schools standardising on digital worksheets; not for generating new content. Pricing: free basic plan; Teacher plan $149 per year; school and district pricing custom (verified 24 August 2026).
Best for newcomers: Glint
Glint builds visual supports, communication boards and social stories from a single prompt, with voice output in 50+ languages, and that makes a newcomer's first weeks dramatically easier: tap a symbol, hear the word in English and in the home language. It was built for special education and speech pathology, and I include it here deliberately. Newcomer English learners are not special education students, but while their English arrives they benefit from the same visual scaffolds. It declares FERPA and COPPA compliance with US hosting. What sets it apart from a generic image generator is that its boards speak, switch languages and print cleanly for a classroom wall. Watch out for the 18+ minimum age, which makes this strictly a teacher tool, and remember a visual board is a bridge, not a curriculum. Best for arrival packs, routines and family sharing; not for language instruction itself. Pricing: free to try; $18 per month or $150 per year; school and district licensing quote-based (verified 24 August 2026).
Best for reading fluency and confidence: EchoGuide Pro
EchoGuide Pro listens while a student reads aloud and guides pace and clarity in real time, with next-word highlighting and a dyslexia-friendly mode. For English learners that low-anxiety rehearsal is gold: reading aloud in a second language in front of peers is one of the most stressful things we ask children to do, and a private practice loop changes the equation. It declares FERPA, COPPA and GDPR compliance, takes students from age 4, and hosts data in the UK, which US districts should note in procurement even though it is not a blocker. What sets it apart from Snorkl is the focus on fluency and delivery rather than on explaining reasoning. Watch out for its youth: the tool launched in 2026, the evidence base is early, and there is no public price list. Best for daily fluency practice from elementary up; not for schools that need a published per-seat price today. Pricing: paid, quote-based, with a 7-day free trial (verified 24 August 2026).
Best for listening practice: Dictation Trainer
Dictation Trainer does one thing well: students listen, type what they hear, and get word-by-word feedback, with lessons graded A1 to B2+ and the option to turn any YouTube video into a listening exercise. For older English learners the CEFR grading maps neatly onto where they are. What sets it apart from the big platforms is honesty of scope; it is a listening gym, nothing more. Watch out before any school rollout: the minimum age is 13+, it collects student data (name, email, performance), GDPR compliance is listed as in progress, and data rides on a global Cloudflare network rather than a stated US home. Treat it as a recommendation for students 13 and over, or run it through your district's data review first. Best for independent listening practice at 13+ and adult learners; not for younger students. Pricing: free tier with daily credits; Pro at $6.99 per month or $59.99 per year (verified 24 August 2026).
How to Choose
The rule I give every school I advise is to outsource the doing, not the thinking, and for English learners it needs a sharper edge, because here the struggle in the language is the learning. So split the job in two. Lower the barrier to input: levelled texts, read-aloud, translated instructions and family letters. Diffit, Kami and Glint should carry as much of that load as they can, because none of it is what the student needs to practise. Then protect the practice of output: speaking, listening, reading and writing in English must stay effortful, and the tool's role there is feedback and a safe place to rehearse. Snorkl, EchoGuide Pro and Dictation Trainer live on that side of the line.
Three rules for the meeting where you decide. If your students are under 13, the COPPA column is your first filter, not your last. If you can only fund one thing, fund the input side; it helps every teacher, in every subject, tomorrow. And if a tool's pitch is that students will not need English because it translates everything, walk away. That is a ceiling dressed as a ramp.
What I would do
Diffit for levelled input, Snorkl for spoken output, Kami to make every document accessible, MagicSchool if you want one governed platform, and Glint, EchoGuide Pro and Dictation Trainer for newcomers, fluency and listening. Only 32% of the 90 tools on aieducator.tools declare COPPA compliance, so read the safety line before the feature list, and if you teach English as a subject rather than to English learners, the English teachers list is the better fit. Your multilingual learners are not a niche. They are one in ten of your students, and they have waited long enough for tools built with them in mind.
Dan Fitzpatrick is a Forbes contributor, three-time bestselling author and founder of The AI Educator, and has trained more than 150,000 educators across 30+ countries.