Key takeaways

• Diffit’s free tier is genuinely unlimited for core generation. Twee’s is a 5 run trial. If cost is the deciding factor, this settles it.
• Only Twee works in CEFR. Diffit uses US reading levels, which is a real mismatch for international English teaching.
• Diffit does one job very well. Twee does many jobs across a lesson, from reading text to gap fill to assessment.
• Verified pricing on 19 September 2026: Diffit individual is 14.99 dollars a month or 149.99 a year. Twee Basic is 6.50 a month billed annually, Pro is 10.50.
• Diffit’s paywall is mostly about export format. Free users get PDF, paid users get Google Docs, Slides and Word.
• Twee is cheaper than Diffit at every paid tier, which surprises most people.

The one line answer

If you have a text and need it at three levels, use Diffit and pay nothing. If you have a topic and need a lesson, use Twee and pay for it.

Diffit vs Twee side by side

Compiled 19 September 2026 from each platform’s official pricing pages.

Where Diffit wins

The free tier is real. This is the strongest argument for Diffit and it is not close. Core generation is unlimited, with no credit card and no generation count. Most edtech free tiers are trials wearing a disguise. This one is not.

It is faster at its one job. Paste a URL, choose your levels, and get every version at once with comprehension questions and vocabulary attached. Twee can produce a levelled text too, but Diffit generates the variants simultaneously, which is the whole point when you have a mixed ability group.

Privacy is simpler. Diffit collects no student data at all, which makes the conversation with a school data protection officer very short.

It works from awkward sources. URLs, PDFs and video, not just typed topics.

Where Twee wins

CEFR, natively. Diffit thinks in US grade reading levels. If you teach a B1 class in Spain or an A2 group online, you are translating between systems every time, and the mapping is approximate at best. Twee just asks which CEFR level you want.

Breadth within the lesson. Diffit gives you a text, questions and vocabulary. Twee gives you those plus dialogues, gap fills, discussion prompts, lead in activities, listening work and assessment, all from the same source.

It is cheaper. This catches people out. Twee Pro at 10.50 dollars a month billed annually undercuts Diffit’s individual plan at 14.99, while covering considerably more ground.

Export is included. Twee exports to PDF, Word, Google Docs and Google Forms on the paid plans. Diffit reserves Google and Word export for paying users and gives free users PDF only.

The reading level problem, in detail

This is the part most comparisons skip, and it matters more than any feature list.

Diffit levels text by US grade reading level, which measures reading ability in a first language English speaker. CEFR measures language proficiency in a second language learner. They are not the same axis. A text pitched at a US fifth grader can contain idiom and cultural assumption that an adult B2 learner finds harder than a text two grades above it.

In practice Diffit output still works for ESL, and plenty of teachers use it happily. But you are eyeballing the fit rather than setting it, and you will occasionally be wrong in ways that only show up in class. Our guide to Diffit for teachers goes into where this bites hardest.

Who should pick which

Pick Diffit if your classes are mixed ability and your main task is making one text accessible to everyone, if you have no budget, or if you teach within a US school system where grade levels are the working currency.

Pick Twee if you teach in CEFR terms, if you build materials from scratch rather than adapting existing ones, or if you want exercises and assessment as well as text.

Use both if you can. Diffit free costs nothing and handles differentiation. Twee Pro at 10.50 a month handles production. Together that is cheaper than a Diffit individual subscription alone.

A fair test before you decide

Take one real lesson you have to teach next week. Run it through both. Not a demo topic, an actual class with an actual level.

With Diffit, check whether the levelled output suits your learners or whether you are mentally adjusting for the grade to CEFR gap. With Twee, check whether the five trial runs get you something you would genuinely put in front of a class.

That test takes half an hour and tells you more than any comparison table, including this one.

Read further

Full detail in the Twee review for English teachers and in What is Diffit?. If neither fits, see Twee vs MagicSchool or the wider guide to the best AI tools for English teachers.

You can also try our free Lesson Plan Generator, which builds a CEFR levelled ESL lesson with no account wall, and read the wider guide to AI in English language teaching.

Frequently asked questions

Is Diffit or Twee better for ESL teaching?

Twee, if you work in CEFR terms. It treats language level as a native setting while Diffit uses US grade reading levels, which measure first language reading ability rather than second language proficiency. Diffit is still useful for ESL, but you are estimating the fit.

Which one is free?

Diffit, meaningfully. Its free tier gives unlimited core generation with PDF export and no credit card. Twee’s free tier is a 5 run trial intended for evaluation rather than teaching.

Which is cheaper to subscribe to?

Twee. Basic is 6.50 dollars a month billed annually and Pro is 10.50, against Diffit’s individual plan at 14.99 a month or 149.99 a year. Twee also covers a wider range of tasks at that price.

Can Diffit generate exercises as well as reading texts?

Yes, but a narrower range. Diffit produces comprehension questions, vocabulary lists and graphic organisers alongside the levelled passage. Twee offers more than 40 tools covering gap fills, dialogues, discussion questions and assessment.

Do I have to choose between them?

No, and the combination is often the best value. Diffit free handles differentiating existing texts while Twee Pro handles building new material. At 10.50 dollars a month total that costs less than subscribing to Diffit alone.

Does Diffit support CEFR levels?

Not directly. It works in US reading levels, roughly second grade through high school. There is no CEFR setting, so matching output to A2 or B1 means judging it yourself rather than selecting it.

Key takeaways

• Nothing replaces Twee feature for feature. Every alternative trades away either the CEFR levelling or the language specific exercise types.
• Most people searching for an alternative are reacting to the 5 run free trial rather than to the product. Twee Basic is 6.50 dollars a month billed annually, which is cheaper than the alternatives people switch to.
• Diffit is the strongest free option, but works on reading level rather than CEFR and only handles text levelling.
• MagicSchool has the most generous free tier in the category, though it has no native CEFR setting.
• A general assistant with a saved prompt preamble gets surprisingly close, at the cost of consistency and export quality.
• Curipod is often suggested here but no longer fits, having moved to US assessment aligned writing practice.

Why the alternatives are all compromises

Twee occupies an unusually narrow position. Almost every other AI teaching tool is built for general school education, where level means a school grade and the subject could be anything. Twee assumes the subject is language and the level is CEFR.

That means the question is not which tool matches Twee, but which part of Twee you actually need. Once you answer that, the choice gets easy.

The alternatives, compared

Compiled 19 September 2026 from each vendor’s official pricing and feature pages.

If you want levelled material for nothing

Diffit is the answer, with the caveat that it levels by reading difficulty rather than CEFR. For a lot of practical classroom work that distinction matters less than it sounds, particularly with younger learners or general reading practice.

Where it shows is precision. If you need a text that is reliably A2 rather than roughly easy, Diffit will not give you that certainty. See Diffit for teachers for the detail.

If you want more than materials

MagicSchool covers lesson planning, rubrics, feedback, report comments and family communication, with a free tier that includes every tool permanently. For a teacher whose main pain is the administrative tail rather than the lesson itself, that is a better use of money than any materials tool.

The trade is the one that sends people to Twee in the first place. No CEFR setting, so levelling drifts unless you prompt carefully every time. The MagicSchool review and the direct comparison both go deeper.

If you resent paying at all

A general assistant with a saved prompt preamble gets closer than most people expect. State the CEFR level, the first language background, the age group and the variety of English you want, paste it every time, and the output is usable.

What you lose is consistency across generations, the exercise structure that arrives automatically in a purpose built tool, and the export formatting. Expect to spend the saved subscription money in time instead. Our comparison of the main ChatGPT alternatives for teachers covers which assistant handles this best, and ChatGPT for English teachers has the prompts worth saving.

Before you switch, check the price again

A lot of people go looking for Twee alternatives because they believe it costs around 12 dollars a month. Several widely read directories publish that figure and it is wrong.

Read on the official page on 19 September 2026: Basic is 6.50 dollars a month billed annually and Pro is 10.50. If price was the reason for looking, the reason may not hold. The full Twee review sets out what each tier includes.

What most teachers end up doing

Not switching. The common pattern is Twee for materials, MagicSchool free for feedback and administration, and a general assistant for anything odd. Between them that covers almost everything, and only one of the three costs anything.

If you want to see what proper CEFR levelling produces before paying anyone, our free Lesson Plan Generator builds a levelled ESL lesson with no account needed. The best AI tools for English teachers guide covers the full landscape, and AI in English language teaching covers where they help across a week.

Frequently asked questions

What is the best alternative to Twee?

There is no direct equivalent. Diffit is the best free option for levelling reading texts, MagicSchool the best for breadth across a teaching week, and Brisk the best for feedback inside documents. Each trades away either CEFR levelling or language specific exercise types.

Is there a free alternative to Twee?

Yes. Diffit offers unlimited core generation free with PDF export, and MagicSchool’s free plan includes all 80 plus teacher tools with capped volume. Neither works in CEFR, so levelling for language learners is less precise.

Why do people look for Twee alternatives?

Usually the 5 run free trial rather than the product itself. Many also believe Twee costs around 12 dollars a month, a figure several directories publish incorrectly. Verified on 19 September 2026, Basic is 6.50 dollars a month billed annually and Pro is 10.50.

Does any alternative support CEFR levels natively?

Not among the mainstream options. Diffit uses United States reading levels, MagicSchool assumes school grades, and general assistants only apply CEFR if you specify it in the prompt. This is the gap Twee fills and the reason switching usually involves a compromise.

Is Curipod a good Twee alternative?

Rarely, now. Curipod has moved from interactive slides to writing practice aligned to United States assessments such as TELPAS, ELPAC and WIDA ACCESS. For teachers working outside American school systems it addresses a different need entirely.

Can a general assistant replace Twee?

Partly, with discipline. A saved prompt preamble stating CEFR level, first language background, age group and variety of English produces usable material. You lose consistency between generations, automatic exercise structure and clean export formatting.

Key takeaways

• Before switching, name the reason. MagicSchool’s weakness for English teachers is CEFR levelling, not breadth, and most alternatives are narrower rather than better.
• Twee is the closest thing to a direct replacement for materials work, at 6.50 dollars a month billed annually for Basic and 10.50 for Pro, verified 19 September 2026.
• Diffit’s free tier is genuinely unlimited for core generation, with PDF export only. Paid is 14.99 dollars a month or 149.99 a year.
• Brisk publishes no clear individual price. Sources variously report free forever, 99.99 dollars a year and quote only, which is reason enough to check the site before budgeting.
• Curipod is now a writing feedback platform aligned to United States assessments, so it is rarely the right move for international English teachers.
• MagicSchool’s free tier is the most generous in the category, so leaving entirely is often the wrong call. Adding a second tool usually beats replacing the first.

First, why are you leaving?

This matters more than the list. MagicSchool has over 80 teacher tools on a free plan that caps volume rather than access, which is unusually generous. Most teachers who abandon it do so for one of four reasons, and only two of them are solved by switching.

No CEFR support. The real one. MagicSchool assumes grade levels because it was built for United States K12, so levelling for A2 or B1 means describing it in every prompt and accepting drift. A language specific tool fixes this.

Materials quality. Related to the above. Generic worksheets need heavy editing for language learners. Also fixed by switching.

Hitting the free cap. Not fixed by switching. Most alternatives have tighter free tiers, not looser ones. Paying 8.33 dollars a month for Plus is usually cheaper than the alternative.

Too many tools. Also not fixed by switching. The catalogue is large but you only need the four or five you actually use.

The alternatives compared

Pricing verified 19 September 2026 from each platform’s official pages, except Brisk, where published figures conflict across sources.

Twee, if you want the materials to be right first time

The strongest alternative for English language teachers specifically. Twee treats CEFR as a setting rather than a prompt, covers more than 40 language teaching tools across ten languages, and exports to PDF, Word, Google Docs and Google Forms.

The trade off is breadth. Twee will not write your report comments or draft a parent email, and its free tier is five runs rather than a usable plan. If materials are where your hours go, it is the obvious move. Full detail in our Twee review, and a direct head to head in Twee vs MagicSchool.

Diffit, if the job is levelling one text

Diffit does one thing and does it cleanly: paste a text, topic, URL or video link and get several reading level versions with comprehension questions and vocabulary. The free tier is genuinely unlimited for that core workflow, with the paywall sitting mostly on export formats.

For English language teaching the caveat is that Diffit works in United States reading levels rather than CEFR, so you are still translating between systems. See Diffit for teachers for where that matters.

Brisk Teaching, if you live in Google Docs

Brisk is a browser extension rather than a platform, layering feedback, levelling and quiz tools over documents you already have open. For marking inside Google Docs it removes more clicks than anything else here.

Be careful with the pricing. Published figures conflict badly: some sources describe a free forever educator tier, others an Educator Pro plan at 99.99 dollars a year, others quote only school plans. Check the official site before committing a budget. Our Brisk Teaching review covers what it does well.

Curipod, only if you teach in a US school

Curipod has moved from interactive slides to writing practice with fast feedback, and its language learner features are built around TELPAS, ELPAC and WIDA ACCESS. Those are United States assessments. If you teach CEFR based courses or prepare learners for IELTS or Cambridge, most of that value does not transfer. Our Curipod review explains the shift.

A general assistant, for everything the platforms miss

Almost every teacher ends up with one open anyway. General assistants handle the unusual requests, the rewrites and the one off problems that no purpose built tool anticipated. They need detailed prompting for level accuracy and they will not flag a culturally awkward example, but they are the most flexible option and the free tiers are workable. See our comparison of the main ChatGPT alternatives for teachers.

What most English teachers should actually do

Not switch. Add.

Keep MagicSchool on the free plan for feedback, rubrics, reports and admin, because those jobs do not depend on language level and the free tier handles them. Then add one language specific tool for the part MagicSchool genuinely cannot do, which is levelled materials.

That combination costs less than most single subscriptions and covers more of a teaching week than replacing one platform with another. Switching outright only makes sense if you never used MagicSchool for the admin side in the first place.

For the wider field, our guide to the best AI tools for English teachers covers how these fit together, and our free Lesson Plan Generator builds a CEFR levelled ESL lesson with no account wall if you want to test the approach first.

Frequently asked questions

What is the best MagicSchool alternative for ESL teachers?

Twee, in most cases. It is the only widely used alternative that treats CEFR level as a native setting rather than something you describe in a prompt, which is precisely the gap English language teachers hit with MagicSchool.

Is there a free alternative to MagicSchool?

Diffit has the strongest free tier among the alternatives, with unlimited core generation and PDF export. Brisk Teaching also offers a free extension. That said, MagicSchool’s own free plan is more generous than either, so leaving for cost reasons rarely helps.

Why does MagicSchool struggle with English language teaching?

It was built for United States K12 general education, so the interface assumes grade levels and the output defaults to American conventions. There is no native CEFR setting, which means levelling for language learners requires prompting every time and still drifts.

Is Brisk Teaching cheaper than MagicSchool?

Hard to say, because Brisk does not clearly publish an individual price. Reported figures range from free forever to 99.99 dollars a year to quote only for schools. MagicSchool publishes 8.33 dollars a month billed annually, which at least makes budgeting possible.

Should I replace MagicSchool or add another tool?

Usually add. MagicSchool’s free tier handles feedback, rubrics and administration well, and those tasks do not depend on language level. Pairing it with one CEFR native tool for materials covers more ground than any single replacement.

Does any alternative handle IELTS or Cambridge preparation?

None of these do it properly. Twee is built around general English, Diffit around United States reading levels, and Curipod around United States state assessments. Exam specific material still needs building by hand or with a general assistant and a detailed prompt.

Key takeaways

• Pick Twee if your bottleneck is producing levelled reading texts, exercises and assessments. Pick MagicSchool if it is marking, reports and everything around the lesson.
• Verified pricing on 19 September 2026: Twee Basic is 6.50 dollars a month billed annually and Pro is 10.50. MagicSchool Plus is 8.33 dollars a month billed annually or 12.99 monthly.
• MagicSchool has by far the better free tier. All 80 plus teacher tools are included and only volume is capped. Twee’s free tier is a 5 run trial.
• Only Twee treats CEFR as a native setting. MagicSchool needs the level spelled out in every prompt and still drifts.
• MagicSchool is the safer institutional buy, with FERPA, COPPA, SOC 2, GDPR and Ed Law 2-D compliance and single sign on at Enterprise level.
• Running MagicSchool free alongside Twee Pro costs 10.50 dollars a month and covers more ground than either subscription alone.

The short answer

If you only read one paragraph: these tools barely overlap. Twee is a materials factory for language teachers. MagicSchool is a swiss army knife for school teachers. Choosing between them on price or tool count misses the point, because they are not trying to do the same job.

The question worth asking is not which is better. It is which part of your week actually hurts.

Twee vs MagicSchool at a glance

Compiled 19 September 2026 from each platform’s official pricing page.

Where Twee clearly wins

Levelling. This is the whole argument. Twee treats CEFR as a parameter rather than a suggestion, so a B1 reading text arrives close to B1. Ask MagicSchool for the same thing and you are describing the level in prose and hoping. Over a term that difference is hours of editing.

Building a whole lesson from one source. Paste an article, pick a level, and generate the text, vocabulary, gap fill, comprehension questions and discussion prompts in one sitting. MagicSchool can produce each of those separately, but they will not cohere around the same source in the same way.

Video into listening work. Twee’s media tools take a video link and produce comprehension material from it. Building that by hand is one of the slowest jobs in language teaching.

Finishing the job. Export to a formatted PDF, Google Form or interactive student link means the material is ready rather than needing reformatting.

Where MagicSchool clearly wins

The free tier. It is not close. MagicSchool gives you every one of its 80 plus teacher tools for nothing and caps only volume and history. Twee gives you five runs. If budget is the constraint, this decides it.

Marking at scale. Class Writing Feedback drafts comments on a whole set of written work against your own rubric, which you then edit and send. For anyone teaching large groups this recovers more time than any materials tool.

Everything that is not a lesson. Report comments, parent emails, newsletters, meeting notes and behaviour records. Dull, unavoidable, and genuinely level neutral, so the lack of CEFR support costs you nothing here.

Buying it through an institution. The compliance list and single sign on matter if there is a procurement process or a data protection officer between you and the purchase.

What neither of them does

Worth saying plainly, because both are marketed as though they do.

Neither knows your learners. Neither can tell you the text is too easy for four students and impossible for two others. Neither reliably catches a culturally awkward example, and both will hand you something confidently wrong if you stop reading.

Neither is built around IELTS or Cambridge. Twee is general English, MagicSchool is US K12. Exam specific material still gets built by hand or with a general assistant and a detailed prompt.

Both default to American conventions in output. Twee less so, but if you teach British or Irish English you will be editing either way.

The combination most teachers should actually run

Because MagicSchool’s free tier is real and Twee’s is not, the sensible setup for most English teachers is not a choice at all.

Run MagicSchool on the free plan for feedback, rubrics, reports and admin. Those jobs are level neutral, so the free volume cap is rarely the binding constraint and you lose nothing by not paying.

Add Twee Pro at 10.50 dollars a month if you build materials weekly. That is where the CEFR advantage compounds and where the money is justified.

Total cost is 10.50 dollars a month, and it covers more of the week than either subscription alone. Upgrade MagicSchool to Plus only if you hit a ceiling you can name, most likely output history or Student Rooms.

If you teach a fixed syllabus from a course book, or only a few hours a week, stay on MagicSchool free and skip Twee entirely.

How to decide in one question

Look at last week. If the hours went into making things that did not exist, choose Twee. If they went into marking, reporting and answering people, choose MagicSchool. If both, you already know the answer.

Whichever way you go, start with the free option and give it one real task rather than a test drive. A tool that saves you an hour on something you actually had to do tells you more than an afternoon of exploring features.

Read the full reviews

This comparison is a summary. For the detail on either platform, including where each one stops being useful, see the full Twee review for English teachers and the full MagicSchool review for English teachers. If neither fits, our guide to the best AI tools for English teachers covers the wider field, including Diffit and Brisk Teaching.

Before paying for either, try our free Lesson Plan Generator, which builds a CEFR levelled ESL lesson with no account wall, and read the wider guide to AI in English language teaching.

Frequently asked questions

Is Twee or MagicSchool better for ESL teachers?

Twee for materials, MagicSchool for everything else. Twee treats CEFR level as a built in setting, which matters enormously for levelled texts and exercises. MagicSchool is stronger on feedback, rubrics and administration, none of which depend on language level.

Which one has the better free plan?

MagicSchool, by a wide margin. Its free plan includes all 80 plus teacher tools permanently and caps only generation volume and output history. Twee’s free tier is a 5 run trial intended for evaluation rather than ongoing teaching.

Can I use both together?

Yes, and most English teachers should. MagicSchool free handles feedback and admin while Twee Pro handles materials, for a combined 10.50 dollars a month. That covers more of a teaching week than either subscription on its own.

Which is cheaper?

Twee, at the entry level. Twee Basic is 6.50 dollars a month billed annually against MagicSchool Plus at 8.33. But MagicSchool’s free plan is genuinely usable, so in practice MagicSchool costs most teachers nothing.

Does MagicSchool support CEFR levels?

Not natively. It was built for United States K12 education and assumes grade levels, so CEFR has to be specified in each prompt. The output is usually reasonable but drifts, which is the main reason language teachers pair it with a CEFR native tool.

Which is better for a whole school?

MagicSchool. Its Enterprise tier adds single sign on, integration with school systems, district dashboards and a custom data privacy agreement, and it lists FERPA, COPPA, SOC 2, GDPR and Ed Law 2-D compliance. Twee offers a Schools plan but a narrower institutional feature set.

Key takeaways

• MagicSchool is the broadest AI platform in education, covering lesson planning, quizzes, rubrics, feedback, presentations and family communication rather than materials alone.
• Verified pricing on 19 September 2026: Free is genuinely free forever, Plus is 8.33 dollars a month billed annually or 12.99 billed monthly, and Enterprise is priced on application for districts.
• The free tier includes all 80 plus teacher tools. What you pay for is unlimited generations, output history, editing and the student tools, not access to the catalogue.
• It is built for K12 general education. There is no native CEFR setting, so levelling for English language learners means prompting for it every time.
• Its real strength for English teachers is everything around the lesson rather than the lesson itself: feedback at scale, rubrics, reports and parent messages.
• Privacy and compliance are unusually well covered, which matters if you teach young learners or work inside an institution with a procurement process.

What MagicSchool actually is

MagicSchool is a platform of small, single purpose AI tools for teachers. Rather than one chat box, you pick the job you need doing and get a tool built around it: a lesson plan generator, a rubric generator, a worksheet generator, a quiz maker, a report comment writer, a family email drafter.

According to MagicSchool’s own pricing page, the platform runs more than 80 teacher tools and over 50 student tools, alongside a chatbot called Raina, batch class writing feedback and quizzes with real time summaries. Enterprise customers get single sign on, integration with school systems and district dashboards.

The scale is the point. Where a tool like Twee goes deep on language teaching specifically, MagicSchool goes wide across everything a teacher does in a week. That trade off decides whether it suits you.

What MagicSchool costs, checked today

These figures were read from the official pricing page on 19 September 2026.

• Free. Zero, described as forever free. Includes all 80 plus teacher tools, the Raina chatbot, quizzes, class writing feedback, basic support and the full privacy and security package. Generation volume and output history are capped.
• Plus. 8.33 dollars per user a month billed annually, or 12.99 billed monthly. Adds unlimited generations, unlimited output history, unlimited output editing through Studio Mode, the 50 plus student tools, Student Rooms, student learning insights and early feature access.
• Enterprise. Custom pricing for schools and districts. Adds a custom data privacy agreement, single sign on, integration with school information and learning management systems, district alignment, tool management controls, dashboards and a dedicated customer success manager.

The free tier deserves more attention than it usually gets. You are not paying for access to the tools, because they are all in the free plan. You are paying for volume and for the student facing side. A teacher generating a handful of things a week may never hit the ceiling.

Where MagicSchool earns its place for English teachers

The honest case for MagicSchool in an English language teaching context is not materials generation. Tools built for language teaching do that better. The case is everything else.

Feedback at scale. Class Writing Feedback lets you review, personalise and send AI drafted comments on a whole set of written work against your own rubric. If you teach large groups and mark writing regularly, this is the single biggest time recovery on the platform.

The administrative tail. Report comments, family emails, newsletters, behaviour notes and meeting summaries are real work that nobody trains you for and nobody enjoys. MagicSchool handles them competently, and unlike materials they do not need to be levelled for a learner.

Institutional credibility. The platform lists FERPA, COPPA, SOC 2, GDPR and Ed Law 2-D compliance, and Common Sense Privacy certification. If you teach in a school with a procurement process or a data protection officer, that matters more than any feature.

Supervised student use. Student Rooms on the paid plans let you configure which AI tools learners can use and watch what happens. For anyone wrestling with an AI policy, that is a more realistic answer than banning it.

Where it falls short for language teaching

MagicSchool was built for K12 general education in the United States, and that shows in ways that matter to an English language teacher.

There is no native CEFR setting. Where a language specific tool treats A2 or B1 as a parameter, MagicSchool treats it as something you mention in a prompt. The output is usually reasonable, but it drifts, and you end up writing the same instruction repeatedly.

Grade level is the default mental model. Tools ask what grade you teach, which is meaningless for a mixed age adult class in a language school, an online one to one learner in Brazil or a group preparing for an exam.

Exam preparation is largely absent. IELTS, Cambridge and TOEFL formats are not built in, so anything exam specific needs constructing from scratch.

The output reads American. Spelling, cultural references and example contexts default to a US classroom, which needs editing for British or Irish English contexts and for learners who will never set foot in one.

The company reports use by millions of educators across many countries. That is company reported rather than independently audited, and it tells you about reach rather than about fit for language teaching specifically.

How MagicSchool compares with the alternatives

The realistic pattern for an English teacher is MagicSchool for the surrounding work and something language specific for the materials. The table sets out why.

Feature comparison compiled 19 September 2026 from each tool’s official pricing and feature pages.

For how these fit together across a week, see our guide to the best AI tools for English teachers. There are also full reviews of Brisk Teaching and Diffit for teachers, plus a comparison of the main ChatGPT alternatives for teachers.

Who should pay, and who should stay free

Stay on the free plan if you teach a few classes a week, use AI occasionally, and mostly want help with report comments and the odd worksheet. The free tier is unusually generous and there is no pressure to upgrade.

Pay for Plus if you mark writing in volume, want the output history so you can find last term’s work, or want supervised student access through Student Rooms. Those three are what the money buys.

Look at Enterprise only if you are buying for a school and need single sign on, a data protection agreement and oversight. Individual teachers gain nothing from it.

Look elsewhere first if your main need is levelled reading material and exercises for language learners. A language specific tool will get you there with less prompting.

Getting the most out of it as an English teacher

Two habits make a large difference. The first is building your own prompt preamble and reusing it. Something that states the learner level in CEFR terms, the first language background, the age group and the variety of English you want. Paste it in every time. It removes most of the drift.

The second is choosing the right jobs for it. Do not fight the platform to produce a perfect B1 reading text when a language specific tool does it natively. Use MagicSchool for the work that is genuinely level neutral: feedback, rubrics, reports, emails, presentation outlines and lesson structure.

Used that way, the free tier alone covers a lot of ground, and the decision to upgrade becomes obvious rather than speculative.

The verdict

MagicSchool is the most complete AI platform in education and probably the safest one to bring into an institution, but it is not a language teaching tool and it does not pretend to be. For an English language teacher it is best understood as the platform that handles everything around the lesson, leaving the lesson itself to something built for CEFR work.

The free tier is the most generous in this category by a distance, which makes the usual advice straightforward. Start free, use it for feedback and administration, and only pay when you hit a ceiling you can actually name.

If you want something built for English language teaching from the start, our free Lesson Plan Generator produces a CEFR levelled ESL lesson with no account wall, and our wider guide to AI in English language teaching covers where these tools genuinely help and where they do not.

Frequently asked questions

Is MagicSchool actually free?

Yes. The free plan is described as free forever and includes all 80 plus teacher tools, the Raina chatbot, quizzes and class writing feedback. What it caps is the number of generations and how much output history you keep. Many teachers never reach that ceiling.

What does the paid plan add?

Plus costs 8.33 dollars a month billed annually or 12.99 billed monthly. It adds unlimited generations, unlimited output history, unlimited editing through Studio Mode, the 50 plus student tools, Student Rooms and student learning insights. You are paying for volume and student access, not tool access.

Is MagicSchool suitable for teaching English as a foreign language?

Partly. It was built for K12 general education in the United States, so there is no native CEFR setting and the interface assumes grade levels. It works well for feedback, rubrics and administration, but a language specific tool handles levelled materials with far less prompting.

Can students use MagicSchool safely?

Student Rooms on the paid plans let a teacher choose which AI tools learners can access and monitor what they do. The platform lists FERPA, COPPA, SOC 2, GDPR and Ed Law 2-D compliance, which matters if you teach young learners or work inside an institution.

How does MagicSchool compare with Twee?

They solve different problems. MagicSchool goes wide across everything a teacher does in a week, while Twee goes deep on language teaching materials. Many English teachers use MagicSchool for feedback and admin and something CEFR native for the lessons themselves.

Does MagicSchool use American English?

By default, yes. Spelling, cultural references and example contexts assume a United States classroom. If you teach British or Irish English, state the variety you want in your prompt every time, or edit the output before it reaches learners.

Key takeaways

• Almost every Curipod review online describes a product that no longer matches the company’s own pricing page. Check the source before choosing it.
• Verified on 19 September 2026: the platform now leads on writing practice with instant feedback, with tiers of Free, School and District, and an ELA Curriculum Alignment plan. Only the free tier has a published price, which is zero.
• The free plan gives a capped number of weekly sessions with unlimited student writers, and students never need to create an account.
• Its language learner support is built around US assessments rather than CEFR, which limits its usefulness in most international teaching contexts.
• Privacy is a genuine strength: no student accounts, no AI chatbots for learners, and no student or teacher data used to train models.
• The paid tiers are quote only, which makes it hard for a freelancer or small school to evaluate without talking to sales.

What Curipod is now

Curipod is a Norwegian company that built its reputation on interactive lesson slides. Teachers generated a deck from a topic, added polls, word clouds, drawing prompts and open responses, and students joined a live session with a code.

That product still exists in the wild, but it is no longer what the company leads with. Curipod’s own pricing page now frames the platform around one thing: students write, get specific feedback in seconds, discuss and revise. The tagline on the page is that every student writes and every student gets feedback.

This matters because the gap between the reviews and the reality is now wide. A teacher who reads a 2026 round up describing Curipod as a slides tool, then signs up expecting that, will find a writing platform organised around a rather different job.

What Curipod costs, checked today

Read from the official pricing page on 19 September 2026.

• Free. Zero. A set number of writing sessions each week that renews automatically, unlimited student writers, standards alignment, limited access to test preparation content, and basic support. Student written responses are capped at 1,000 characters.
• School and District. Custom pricing, quote only. Unlimited sessions, full test preparation libraries, bilingual writing practice, text to speech, rubric scoring against your own criteria, 5,000 character responses, PDF download, growth reporting, priority support and a signed data privacy agreement.
• ELA Curriculum Alignment. Custom pricing. Adds writing tasks mapped to a district’s adopted curriculum and pacing.

There is no published per teacher price for anything above free. Curipod sells through schools and districts, and the company lists discounts for small schools under 200 students, financial aid arrangements and help finding grants. That is a fair approach for institutions and an awkward one for an individual teacher trying to compare costs.

Where Curipod is genuinely good

Writing volume with feedback attached. The core loop is sound teaching. Students write regularly, get specific feedback quickly, discuss it and revise. Most teachers know that is what improves writing and most teachers cannot sustain it by hand across full classes.

No student accounts. Learners join without signing up. For anyone who has tried to get thirty teenagers through a registration process, or who works in a context with strict rules about children’s data, this removes a real obstacle.

A serious position on AI safety. The company states that any AI is teacher controlled, that there are no AI chatbots for students, and that student and teacher data is never used to train models. The platform lists FERPA, COPPA, SOPPA and GDPR compliance. Very few tools in this category are that explicit.

Authentic writing controls. Paid plans can disable pasting and third party assistance, so you see what a learner can actually produce. For anyone worried about AI written homework that is a more practical answer than detection software.

Why it fits English language teaching less well than it looks

Curipod does have language learner support, and it is worth being precise about what that support actually is.

The bilingual writing practice on the paid plans is built around TELPAS, ELPAC and WIDA ACCESS. Those are United States English language proficiency assessments used in American school systems. They are not CEFR, not IELTS, not Cambridge. If you teach English in Spain, Vietnam, Ireland or online to adults anywhere, that feature does not apply to your learners.

The same holds for the test preparation libraries, which are aligned to US state tests, and for the curriculum alignment tier, which maps to adopted American curricula.

The published case studies report improvements on US state assessments. Those are company published rather than independently audited, and even taken at face value they tell you about American school outcomes rather than about language learners preparing for an international exam.

Age range is the other constraint. The platform is built for K12. An adult class in a language school, a business English group or a one to one online learner sits outside the design.

How Curipod compares

Here is where each of these tools genuinely fits, so you can rule Curipod in or out quickly.

Feature comparison compiled 19 September 2026 from each tool’s official pricing pages.

For the full picture, see our guide to the best AI tools for English teachers, plus reviews of Brisk Teaching and Diffit for teachers, and a comparison of the main ChatGPT alternatives for teachers.

Who Curipod actually suits

A strong fit if you teach English or ELA in a US school, especially one preparing learners for state assessments, and particularly if your school can buy a site licence.

Worth trying free if you teach school age learners anywhere, want to build a regular writing habit, and value the no account and no chatbot approach to student safety. The free tier costs nothing to test.

Probably not for you if you teach adults, run one to one online lessons, prepare learners for IELTS or Cambridge, or work in CEFR terms. The platform is not built around any of that.

Using it well if you do adopt it

Treat the weekly session allowance as a planning constraint rather than an annoyance. One well prepared writing session a week with real feedback and a revision stage beats three rushed ones, and the cap quietly pushes you toward that.

Write your own rubric rather than relying on a default. The feedback quality depends on what you tell it to look for, and a generic rubric produces generic comments that learners learn nothing from.

Build in the discussion step. The feedback arrives in seconds, which tempts everyone to move on. The learning happens when a student reads the comment, talks about it and rewrites.

The verdict

Curipod is a well built platform with an unusually principled position on student data and AI safety, and its core writing loop reflects how writing actually improves. For a US school teaching ELA it is a serious option.

For most English language teachers reading this, it is not the tool it is often described as. The slides product that made its name has given way to a writing platform tied to American assessment frameworks, and the features aimed at language learners are built for US school systems rather than international ones. The free tier is worth thirty minutes of your time to judge for yourself, but go in knowing what it is now rather than what a directory listing says it was.

If you need something built around CEFR from the start, our free Lesson Plan Generator produces a levelled ESL lesson with no account needed, and our guide to AI in English language teaching covers where these tools help across a full week.

Frequently asked questions

Is Curipod still a slides tool?

Not primarily. Curipod built its reputation on AI generated interactive lesson slides, but the company now leads on writing practice with instant feedback. Most reviews online still describe the older product, which is why signing up can be confusing.

How much does Curipod cost?

The free plan costs nothing and gives a capped number of writing sessions each week with unlimited student writers. The School and District plans and the ELA Curriculum Alignment tier are quote only, with no published per teacher price.

Do students need an account?

No. Students join a session without creating an account, which removes a common obstacle in schools with strict rules about children’s data. Curipod also states there are no AI chatbots for students and that student data is not used to train models.

Is Curipod useful for teaching English outside the United States?

Less than most reviews suggest. Its language learner support is built around TELPAS, ELPAC and WIDA ACCESS, which are United States English proficiency assessments. It is not aligned to CEFR, IELTS or Cambridge, so much of that value does not transfer.

Can Curipod stop students using AI to write their answers?

Partly. Paid plans can disable pasting and third party assistance during a writing session, so you see what a learner produces unaided. That is a more practical approach than detection software, which cannot prove authorship.

Who is Curipod best for?

Teachers of English or ELA in United States schools, particularly those preparing learners for state assessments and able to buy a site licence. For adult classes, online one to one teaching or CEFR based courses, a tool built for language teaching will fit better.

Key takeaways

• Twee is one of the few AI tools built for language teaching rather than adapted to it, so CEFR levels and exercise types are native rather than something you prompt for.
• Verified pricing on 19 September 2026: Trial gives 5 runs, Basic is 6.50 dollars a month billed annually, Pro is 10.50 dollars a month billed annually, and a separate Schools plan covers multiple teachers.
• Most third party listings quote Twee’s pricing incorrectly, some by nearly double. Check the official page before budgeting.
• The free trial is genuinely a trial at 5 runs, not a usable free tier. Treat it as an evaluation, not a plan.
• Twee produces material fast. It cannot tell you whether that material suits the eleven people in your Tuesday evening class, and it will not flag a culturally awkward example.
• If your work is exam preparation, one to one tutoring or business English with a narrow syllabus, the saving is smaller than the marketing suggests.

What Twee actually is

Twee is a web platform that generates language teaching material. You give it a starting point, which can be a topic, a list of vocabulary, a piece of text or a video link, choose a CEFR level between A1 and C2, and it returns something you can put in front of a class.

The catalogue is broad. There are tools for reading texts built around your own vocabulary list, discussion questions, dialogues, gap fills, word and definition matching, open questions, pros and cons lists and lead in activities. According to Twee’s own site the platform covers more than 40 tools across ten languages, with output exportable as PDF, Word, Google Docs or Google Forms, or shared with students as an interactive link.

That last part matters more than it sounds. A lot of AI tools stop at generating text and leave you to format it. Twee finishes the job, which is where the actual time saving sits for most teachers.

What Twee costs, checked today

Pricing is where most reviews of Twee go wrong, so these figures were read directly from the official pricing page on 19 September 2026 rather than taken from a directory listing.

• Twee Trial. Free. 5 runs on any tool except the Pro tools, 500 audio credits, free library resources, unlimited interactive sharing and classroom management.
• Twee Basic. 6.50 dollars a month billed annually. 100 runs a month across all tools except Pro, 2,500 audio credits a month.
• Twee Pro. 10.50 dollars a month billed annually. All Pro tools, unlimited runs on the basic tools subject to fair use, 100 media runs a month, 5,000 audio credits, AI checking of student responses, a CEFR level checker, audio and video to text conversion, and export to PDF, Google Forms and Google Docs.
• Twee Schools. Annual, priced on application, covering multiple teachers with 7,000 audio credits and personal support.

Several widely read AI tool directories currently list Twee at around 11.95 dollars a month, which is close to double the annual Basic price. If you are building a department budget, go to the pricing page itself. Note also that the headline figures are the annual rates. Monthly billing costs more.

The word unlimited carries an asterisk on every plan, pointing at a fair use policy. In practice that is unlikely to bite a single teacher, but a school running heavy daily generation should read it before committing.

Where Twee is genuinely strong

The clearest advantage is that Twee was designed for language teaching rather than adapted to it. When you ask a general assistant for a B1 reading text you are asking it to interpret what B1 means, and the result drifts. Twee treats level as a setting rather than a suggestion, and the output lands closer to the mark more often.

Three things it does particularly well.

Building a full lesson’s worth of material from one source. Paste an article, pick a level, and you can generate the text, the vocabulary list, a gap fill, comprehension questions and discussion prompts in one sitting. That is the single biggest time drain in most teaching weeks and Twee removes a large part of it.

Turning video into a lesson. The media tools take a video link and produce comprehension work from it. Anyone who has tried to build a listening lesson around an authentic clip by hand will recognise how much work that removes.

Finishing the job. Export to a properly formatted PDF, a Google Form or an interactive student link means the material is ready to use rather than ready to reformat. On Pro you can add your own logo, which matters for freelancers and small schools.

Where Twee stops being useful

Twee generates material. It does not teach, and the gap between those two things is where the honest limits sit.

It does not know your learners. It cannot tell you that the gap fill is too easy for the four strongest students and impossible for the two who joined in March. That judgement stays with you, and no level setting replaces it.

It does not reliably catch cultural awkwardness. Generated example sentences and discussion questions occasionally land badly with particular groups, and the tool has no way of knowing that. Read everything before it reaches a class.

Exam preparation is thinner than general English. If your week is mostly IELTS or Cambridge work, a generic B2 reading text is less useful than material built around the actual task format and marking criteria. Twee will help with vocabulary and practice, but it is not built around exam rubrics.

The free trial is not a free plan. Five runs is enough to decide whether you like it and nothing more. Tools with a genuinely usable free tier, such as Diffit, serve a different need if budget is the constraint.

The company reports 6 million exercises created in the past year and more than 500,000 teachers having used the platform. Those are company figures rather than independently audited ones, and they are worth reading as scale indicators rather than proof of quality.

How Twee compares with the other tools English teachers use

Twee sits in a crowded field, and the honest answer is that most teachers end up using two tools rather than one. The table below sets out where each fits.

Feature comparison compiled 19 September 2026 from each tool’s official pricing and feature pages.

For the wider picture, our guide to the best AI tools for English teachers covers how these fit together across a teaching week. There are also full reviews of Brisk Teaching and a comparison of the main ChatGPT alternatives for teachers.

Who should pay for Twee, and who should not

Worth paying for if you teach general English across mixed levels, plan several distinct lessons a week, and currently build materials from scratch. That is the profile where the annual Pro price pays for itself in the first fortnight.

Probably worth it if you are a freelancer or online tutor who needs branded, exportable material and values the interactive sharing. The logo option on Pro is a small thing that reads as professional to private clients.

Probably not worth it if you teach a fixed syllabus from a course book with a full resource pack, if your work is almost entirely exam preparation against a set format, or if you teach only a handful of hours a week. In those cases a free tool plus a general assistant covers the gaps.

Getting a usable result in ten minutes

Most people who try Twee and abandon it do so because they treat it as a vending machine. A better first session looks like this.

Start with a lesson you already know well and would normally build by hand. Pick the level deliberately rather than accepting the default. Generate the reading text first and read it properly before generating anything else, because everything downstream inherits its problems. Then build the vocabulary and exercises from that text rather than from the topic, so the material stays coherent. Export the result, then edit it. The edit is not a failure of the tool, it is the part where your teaching goes in.

If you want a second opinion on the level of what came out, the Pro plan includes a CEFR level checker, which is a more useful feature than its low billing suggests.

The verdict

Twee is the most complete AI materials tool currently built for language teaching specifically, and for general English teachers planning several lessons a week it is one of the few subscriptions that clearly returns its cost. The pricing is lower than most of the internet claims, the export options are better than its competitors, and CEFR levels are treated as a real setting rather than a prompt.

What it is not is a replacement for knowing your class. It produces confident, well formatted material that still needs a teacher to decide whether it is right for the people in the room. Used as a first draft engine it is excellent. Used as an autopilot it will eventually put something in front of a class that should not have got there.

If you want to try the approach without a subscription first, our free Lesson Plan Generator builds a CEFR levelled ESL lesson at no cost and with no account wall, and our wider guide to AI in English language teaching covers where these tools help across a full teaching week and where they clearly do not.

Frequently asked questions

Is Twee free?

Not really. Twee offers a free trial of 5 runs on any tool except the Pro tools, plus 500 audio credits. That is enough to evaluate the platform but not to teach with. Paid access starts at 6.50 dollars a month on the Basic plan when billed annually.

Is Twee better than ChatGPT for English teaching?

For generating levelled materials, usually yes. Twee treats CEFR level as a built in setting rather than something you describe in a prompt, so output lands closer to the target level more consistently. For flexibility and unusual requests, a general assistant is still more capable.

Does Twee work for IELTS or Cambridge preparation?

Only partly. Twee is built around general English rather than exam rubrics, so it helps with vocabulary, reading and practice activities but does not generate material matched to specific exam task formats or marking criteria. Exam specific work still needs building by hand.

Can Twee mark student work?

The Pro plan includes AI checking of student responses and a CEFR level checker. Treat both as a fast first pass that flags patterns rather than a final mark. The judgement about what matters for a particular learner stays with the teacher.

How many tools does Twee have?

Twee lists more than 40 tools covering reading texts, dialogues, gap fills, discussion questions, vocabulary matching and lead in activities, working across ten languages. Output can be exported as PDF, Word, Google Docs or Google Forms, or shared with students as an interactive link.

Is Twee worth it for a part time teacher?

Usually not. If you teach a handful of hours a week, follow a course book with a resource pack, or work mainly on exam preparation, a free tool combined with a general assistant will cover the gaps. The subscription pays off when you plan several distinct lessons a week.

Key takeaways

• In the British Council study of 1,348 English language teachers across 118 countries, 57 per cent use AI to create materials and 53 per cent use it to help learners practise, while 24 per cent still use no AI tools at all.
• Only 20 per cent of those teachers feel sufficiently trained to use AI, against 54 per cent who feel inadequately trained. That is the single widest gap in the evidence base.
• Gallup found that teachers using AI at least weekly recover about 5.9 hours a week, which works out at roughly six additional working weeks across a school year.
• 51 per cent of English teachers disagree that AI could teach English without a teacher by 2035, and the most common reason given is the human element of the classroom.
• Assessment is where regulation bites first. The EU AI Act classifies systems that evaluate learning outcomes as high risk, with obligations that carry documentation, accuracy disclosure and human oversight requirements.
• The realistic forecast is a co-teaching model: AI drafts, the teacher decides, and the teacher’s judgement becomes the scarce and valuable part of the job.

What does the evidence actually say about AI use in English teaching?

The strongest sector specific dataset comes from the British Council, which surveyed 1,348 English language teachers in 118 countries for its study Artificial Intelligence and English Language Teaching: Preparing for the Future. It is the closest thing the profession has to a census of how English teachers actually use these tools rather than how vendors say they do.

The task breakdown is revealing. Creating materials leads at 57 per cent, helping learners practise English follows at 53 per cent, and creating lesson plans sits at 43 per cent. Eighteen per cent of respondents said they used AI for none of the listed tasks. On tools rather than tasks, language learning apps lead at 48 per cent, language generation AI at 37 per cent and chatbots at 31 per cent, with 24 per cent reporting no use of any listed AI tool.

Two things follow from that. First, English teaching sits at the materials preparation end of AI use rather than the instruction end. The tools are doing the work that happens before the lesson, not during it. Second, the profession is not uniformly adopting. Roughly a quarter of English teachers are outside this entirely, which is a much larger holdout group than the general education coverage tends to suggest.

How much time is AI genuinely saving?

The clearest time data comes from outside ELT. Gallup, surveying 2,232 United States public school teachers through the RAND American Teacher Panel, reported that six in ten teachers used an AI tool during the school year and 32 per cent used one at least weekly. Teachers in that weekly group estimated savings of about 5.9 hours a week, roughly six working weeks across a year.

That figure deserves a caveat that most coverage drops. It is self reported and it applies only to weekly users. Occasional users report far smaller gains, and the teachers who report the largest savings are also the teachers who already knew how to prompt well. The time dividend is not automatic. It is a return on a skill that most teachers have not yet been taught.

The United Kingdom picture shows how fast the curve moved. The Department for Education’s Generative AI in Education report recorded teacher use rising from 17 per cent in April 2023 to 42 per cent by November of the same year. A near tripling inside seven months is not a trend, it is a step change, and the professional development response has not kept pace with it.

Why is training the real exposure, not replacement?

The British Council numbers on confidence are the most important finding in the whole evidence base, and they get the least attention. Only 20 per cent of English teachers feel sufficiently trained to use AI in their teaching. Fifty four per cent feel inadequately trained. A further 27 per cent sit neutral, which in survey terms usually means uncertain rather than content.

Read that against the adoption figures and the problem becomes obvious. Most English teachers are already using AI in their professional work while telling researchers they do not feel equipped to do so. That is not a technology adoption gap. It is an unsupervised practice gap, and it is where the genuine risk sits: wrongly levelled material, invented example sentences, cultural assumptions passed to learners unchecked, and student data typed into consumer tools with no data agreement behind them.

UNESCO reached the same conclusion from a policy direction. Its Guidance for Generative AI in Education and Research warns that the release of new tools is outpacing national regulation, which leaves institutions unprepared to validate what they are deploying and user data insufficiently protected. Teachers are being asked to make judgement calls that their employers have not written policy for.

What changes first when AI marks the writing?

Assessment is where the future arrives with legal force attached. The EU AI Act, Regulation 2024/1689, classifies AI systems used to evaluate learning outcomes in education and vocational training as high risk under Annex III. That classification carries real obligations: documented human oversight, disclosed accuracy levels including the conditions under which performance degrades, and logging sufficient to reconstruct a decision after the event.

For an English teacher this is less abstract than it sounds. If a tool produces a CEFR band or an IELTS estimate that affects a learner’s placement, progression or certification, someone has to be able to show how that judgement was reached and who signed it off. A teacher using a band estimator to sharpen their own marking is in a very different position from an institution letting software set a grade.

The practical consequence is that the teacher moves from marker to reviewer of machine judgement, and that is a genuinely different professional skill. It requires knowing where the tool is reliable, where it drifts, and how to overrule it defensibly. If you want to see how that works on real learner writing, the free CEFR writing grader is a useful place to test your own agreement rate against a model’s.

Four shifts, and what the evidence says about each

Figures are as reported by the named sources and are approximate where surveys round. Full sourcing sits in the group’s industry report on the future of AI in English language teaching.

Will AI teach English without a teacher?

The profession’s own answer is no, and it is not a sentimental one. When asked whether AI could teach English without a teacher by 2035, 51 per cent of the British Council respondents disagreed, with the human element of teaching cited most often as the reason.

The technical case behind that view is straightforward. A model can produce fluent, correct English on demand. What it cannot yet do reliably is notice that a learner has gone quiet because they were embarrassed by a correction, decide that today’s plan should be abandoned, or judge which of six errors in a paragraph is the one worth teaching to this learner this week. Those are the decisions that separate a lesson from a worksheet, and they all depend on knowing the person in front of you.

That said, the honest version of this argument includes its weak point. Machine conversation practice is improving quickly, and for pure fluency repetition it is already good enough for many learners. Our own look at AI as a speaking partner sets out where it holds up and where it fails.

What about the varieties of English problem?

This is the finding most likely to matter in the long run and the one least discussed. Eighty per cent of the British Council respondents pointed to the need for AI to support the learning of different varieties of English around the world, to improve inclusivity and realism, and to reduce the pull towards standardisation.

The concern is concrete. Large language models are trained on the English that dominates the internet, which skews heavily towards a small number of standard varieties. A learner in Manila, Lagos or Mumbai who will use English mostly with other non native speakers is being modelled on a form of English that may not reflect their actual communicative reality. A teacher who understands their learners’ English environment is currently the only correction to that, which makes local knowledge a professional asset rather than a limitation.

What should an English teacher do about this now?

The evidence points to three moves, in order of return.

Close the training gap first. Being in the 20 per cent who feel competent rather than the 54 per cent who do not is the highest value change available, and it is the one thing entirely within a teacher’s control. The AI-Skilled Teacher Certificate exists for exactly that, and it is designed to be evidence of competence rather than attendance.

Second, move the time saving into preparation rather than out of the working week. Teachers who use the recovered hours to plan better lessons see a compounding benefit. Teachers who use them to take on more classes see none. Building one repeatable workflow, such as a reliable prompt pattern in the lesson plan generator, beats collecting subscriptions to ten tools.

Third, write down your own rules before an institution writes them for you. What goes into a public tool and what does not. Which judgements stay human. How you handle work you suspect was generated, which is a harder problem than the detection vendors admit, as our piece on spotting AI written work sets out.

The five year outlook

Pulling the evidence together, the most defensible forecast is unglamorous. AI settles in as a standard co-teacher for planning, materials and first pass feedback, in the way that a coursebook and a projector did before it. Assessment tightens rather than loosens, because regulation is already pointed at it. The teacher’s role concentrates on the parts that require knowing a learner, and professional value shifts from knowing the language to judging the output.

The risk in that picture is not unemployment. It is a two tier profession, where teachers with documented AI competence and a defensible workflow are treated as professionals, and teachers without it are treated as interchangeable operators of somebody else’s software. The 54 per cent figure is the warning, and it is the one worth acting on.

If you want the fuller market and industry picture behind this analysis, including the jobs and assessment evidence, read the group’s Future of AI in English Language Teaching and TEFL report. For the practical classroom side, our complete guide to AI in English language teaching covers the workflows in detail.

What the two terms actually measure

Teacher talking time and student talking time are observation categories before they are anything else. They were designed to be counted by somebody sitting at the back of a room, which is why they are coarse: they record who is speaking and nothing about whether the speaking is any good. The table below sets out what each term covers and what it leaves out.

The British Council’s teaching knowledge entry puts the definition plainly: teacher talking time is the time that teachers spend talking in class, rather than learners, and student talking time is the time learners spend talking rather than the teacher. Both, it notes, are useful categories for observation of teaching or for self reflection about teaching. That is the honest scope of the pair. They are lenses, not verdicts.

How it is used in practice

The arithmetic that made TTT a standard training topic is simple. Steve Darn, writing for the British Council, puts it this way: if the teacher talks for half the time in a 60 minute lesson with 15 students, each student gets only 2 minutes to speak. Nobody learns to swim in two minutes a week, and that is the whole argument in one line.

The argument on the other side is equally real. The same source lists the forms of teacher talk that earn their place: personalised presentations, questioning, natural conversation and storytelling. In a monolingual classroom the teacher may be the only source of natural spoken English in the room, and roughly tuned teacher speech is comprehensible input rather than dead air. Silence is also doing work. Darn notes that it provides processing time between instructions, during explanations, while waiting for a student to respond, and during monitoring of activities, and that prompting or rephrasing is often counterproductive when the learner simply needs a moment.

So the practical use is diagnostic, not prescriptive. Declan Cooley, also writing for the British Council, describes the problem as leakage rather than volume: unhelpful teacher talk can leak out in many small, often unnoticed, ways, and he names four of them, which are repeating instructions, saying far more than the learner when receiving a contribution, asking long and elaborately polite questions, and echoing answers the class has already heard. None of those four is a decision a teacher makes. They are habits, which is why measuring is the first step and why Cooley says the first step to reducing teacher talk is simply being aware of it.

Reducing teacher talk is mostly a matter of changing the interaction pattern rather than exercising restraint. Group work with the teacher monitoring raises learner output at a stroke, because several learners speak at once. So does eliciting instead of presenting, demonstrating an instruction rather than narrating it, and cutting the echo. The stage where teacher talk reliably balloons is the presentation stage of a presentation, practice, production lesson, which is one of the standing criticisms of that shape.

A worked example: coding one recorded lesson

A teacher records a 60 minute upper intermediate general English lesson, fifteen learners, and listens back with a timer, marking every stretch of teacher talk into one of four categories. The total came to 26 minutes, or 43 per cent, which by the usual rule of thumb is unremarkable. The breakdown is where the lesson is.

The finding that matters is the last row, and it is not the one the percentage would have surfaced. Forty three per cent is a defensible total; four minutes of teacher talk inside a twelve minute discussion is not, because the discussion was the part of the lesson where the learners were supposed to be doing the work. This is the argument for coding the talk by category and by position in the lesson, rather than reporting one figure. A teacher told to talk less will usually trim the anecdote, which was the useful part. A teacher shown this table stops monitoring out loud, which costs nothing and returns four minutes of learner speech immediately.

Commonly confused

The confusion that causes most damage is the second one. Once a ratio becomes a target, teachers optimise for the target, and the easiest way to cut teacher talk is to cut the explanation and the anecdote, which were the parts a learner could not get anywhere else. The glossary entries on drilling and the affective filter describe the same trap from other angles: a measurable proxy quietly replaces the thing it was standing in for.

Where the idea came from, and the argument about it

The measurement predates the terminology. Interaction analysis in the 1960s, counting classroom moves with an observation schedule, established that teachers do most of the talking. Chaudron’s 1988 review of second language classroom research reports the first language finding that teachers tend to do most of the talking, about 60 per cent of the moves, and concludes that the language classroom studies confirm the standard finding of roughly two thirds of classroom speech being attributable to the teacher. Sinclair and Coulthard had already supplied the structural explanation in 1975: classroom discourse runs on an initiation, response, feedback exchange in which the teacher holds two of every three turns by design.

The terms TTT and STT themselves belong to communicative language teaching and to the initial training certificates that spread it. If the goal is communicative competence, and competence is built by using the language, then a lesson in which learners barely speak is failing on its own terms. Swain’s output hypothesis in 1985 gave that intuition a theoretical footing by arguing that producing language, not only receiving it, is what pushes a learner to process grammar rather than meaning alone.

Then the correction arrived. Thornbury in 1996 had teachers analyse their own classroom talk and found the useful variable was not volume but the kind of question asked and the kind of response invited. Cullen in 1998 made the case most directly: the profession had turned a reasonable observation into a slogan, and the right criterion for teacher talk is whether it is communicative in that classroom, not whether there is little of it. Walsh took the argument furthest, showing in 2002 that teacher talk can construct or obstruct learner participation depending on how it is used, and building a self evaluation framework so that teachers could examine their own talk against the aim of the moment rather than against a percentage.

Where that leaves a working teacher is unglamorous and sound. The figure is worth knowing once, because almost every teacher who measures it is surprised. After that the question is which category the minutes fall into and where in the lesson they land. A teacher who has never counted is probably talking more than they think; a teacher chasing a ratio is probably cutting the wrong minutes. Every technique named on this page has its own entry in the TEFL teaching methods and classroom techniques glossary.

Sources

Five of the sources below carry a link, and each of those five was fetched and confirmed live on 18 September 2026. The rest are cited bibliographically without a link, because this site does not publish links it has not checked.

Frequently asked questions

What is a good ratio of teacher talking time to student talking time?

There is no figure that holds across every lesson, and the ones in circulation, usually 30 per cent teacher and 70 per cent learner, are training heuristics rather than research findings. A beginners class needs more modelling than an advanced discussion class. The ratio is useful as a prompt to look at a lesson, and misleading as a standard to be marked against.

Is teacher talking time always a bad thing?

No. The British Council lists personalised presentations, questioning, natural conversation and storytelling as forms of teacher talk that earn their place, and in a monolingual classroom abroad the teacher may be the only source of natural English a learner hears. What displaces learning is unplanned talk: echoing answers, repeating instructions and filling silence.

How do I measure my own teacher talking time?

Record one lesson on a phone and listen back with a timer, marking each stretch of teacher talk as instruction, model, feedback or filler. The category matters more than the total. Most teachers find the overall figure is defensible and the proportion of filler is not, which is a far more fixable finding than being told to talk less.

What is the difference between teacher talking time and teacher talk?

Teacher talking time is a quantity, measured in minutes or as a percentage of the lesson. Teacher talk is the language itself and its quality, including how instructions are worded, how questions are framed and how feedback is given. Cullen’s argument in 1998 was that the second matters more than the first, and that a good lesson is not identified by a stopwatch.

How do I increase student talking time?

Change the interaction pattern rather than trying to speak less. Pair and group work with the teacher monitoring raises learner output immediately, because several learners speak at once instead of one at a time. Demonstrating an instruction rather than explaining it, eliciting rather than presenting, and waiting several seconds after a question all help.

Does teacher talking time work differently in online lessons?

Yes, and mostly against the teacher. Overlapping speech is harder to manage on a video call, so learners wait for a gap and the teacher fills it. In a one to one online lesson the split is naturally closer to even than in a class of fifteen, so the figure to watch is not the ratio but whether the learner is producing extended turns or only answering questions.

Related entries

Next in the glossary: presentation, practice, production, whose first stage is where teacher talk most reliably expands, and the affective filter, which explains why a silent class is not always a class with nothing to say. The full TEFL teaching methods and classroom techniques glossary holds every entry. For the qualifications and career terminology that sits alongside the teaching vocabulary, see the Institute’s TEFL qualifications and terminology explained, and for where this sits inside a wider method, its guide to the communicative approach in ESL teaching. Teachers working online will find the balance shifts again, which the Irish institute’s guide to starting out online lists among the first habits to correct.

Last updated: 18 September 2026 · Written by: Tara Bourke, tefl.ai · Review cycle: definitions are re-checked against their cited sources every six months.

Key takeaways

• Listening is the skill most teachers under resource, because suitable audio is hard to find and slow to make.
• AI is strongest at drafting scripts and questions, and weakest at producing speech that sounds like people actually talking.
• Always give the model a level, a context, a number of speakers and a reason for the conversation. Vague prompts produce radio announcements.
• Synthetic audio has no overlap, no false starts and no background noise, so it is easier than real life. Plan for that gap.
• Use generated audio for controlled practice and authentic audio for the messy parts. Both belong in a course.
• Recording your own learners is personal data. Get permission before it goes anywhere near a tool.

Why listening is the skill that gets skipped

Most teachers do not avoid listening because they think it is unimportant. They avoid it because the logistics are miserable. The coursebook audio is about a topic your class finished two weeks ago. The podcast you like is far too fast. The news clip is three minutes of vocabulary nobody needs. Making your own recording means writing a script, finding a second voice, recording it without a dog barking, then writing the questions.

That is the entire reason listening lessons often become reading lessons with the teacher performing the text out loud.

Generated audio changes that arithmetic. You can produce a two minute conversation between two speakers, at roughly the right level, about the thing your class is studying this week, in the time it takes to make tea. Whether it is any good is a separate question, and the honest answer is that it depends on what you ask for and what you do afterwards.

What AI can genuinely produce for a listening lesson

A graded script. This is the strongest use by a distance. Ask for a conversation at a stated level, with a stated number of speakers, a stated setting and a stated outcome, and you will get a workable first draft.

Audio from that script. Text to speech has moved far enough that a two voice dialogue sounds like two people rather than a train station announcement. It still sounds like two very calm people.

Comprehension questions. Gist questions, detail questions and inference questions, generated from the script rather than from the recording, which means they are always accurate to the text.

A transcript. You already have it, because you wrote the script. That sounds obvious until you remember how much class time is lost hunting for the line where the speaker said the thing.

Variations. The same situation at two levels, or the same dialogue with one detail changed, which is the basis of several good tasks further down this article.

The prompt that actually works

Most disappointing listening material comes from a prompt like “write a B1 conversation about travel”. What you get back is two people taking turns to deliver complete, grammatical sentences about how much they enjoy travelling. Nobody has ever spoken like this.

Give the model six things instead: the level, the setting, the relationship between the speakers, the reason the conversation is happening, the length in words, and the language you want to appear. Then add one instruction that changes everything: tell it that the speakers should interrupt each other, that one of them should not understand something the first time, and that the conversation should end without everything being resolved.

Add the target language explicitly. If the lesson is about polite requests, say that polite requests must appear at least four times in natural positions. Models are good at following that instruction and terrible at guessing it.

Level bands here are typical starting points rather than fixed rules. Your class will move them.

The problem nobody mentions: it is all too easy

Generated speech is articulate. Speakers wait their turn, finish their sentences, pronounce every word and never say “sorry, what?”. Real English does none of this. People talk over each other, abandon half their sentences, swallow function words and repair what they just said.

A learner who only ever practises on synthetic audio gets very good at understanding synthetic audio. Then they arrive at a service desk in Manchester and understand nothing, which is demoralising and entirely predictable.

The fix is not to abandon generated audio. It is to be deliberate about the split. Use generated material for controlled practice, for the target language, for the things where you need a clean example. Use real speech, with all its mess, for training the skill of coping. Both are necessary, and the second one is where the confidence comes from.

You can also push the model towards mess. Ask for hesitation, self correction, an interruption and one speaker who is distracted. It will not be perfect, but it will be better than the default.

Six listening activities worth the preparation

1. Two versions, one situation. Generate the same conversation twice, changing three facts. Half the class hears one version, half hears the other. They then compare in pairs and find the differences, which forces both groups to talk about what they actually heard rather than what they assumed.

2. Predict the ending. Play the first two thirds. Learners write what happens next, then hear it. Generated dialogues are ideal here because you control where the break falls.

3. Dictogloss. A short dense paragraph, read twice at normal speed, reconstructed in groups. Ask the model for a paragraph built around one grammatical structure and keep it to about eighty words, because the first draft will be double that.

4. The unhelpful speaker. Generate a service conversation where one speaker is vague, distracted or mishears. Learners note what information is missing. This is the closest a generated task gets to real listening.

5. Transcript repair. Take the script, introduce ten errors, and have learners correct it while listening. It sounds fiddly and takes two minutes to make.

6. Extensive listening at home. Generate a short serialised story, one episode per week at a steady level. Learners listen without a task, for pleasure and volume. This is the activity most likely to move a class and least likely to be assigned.

Getting the level right

Level in listening is not just vocabulary. It is speed, density, how much is said indirectly, and how much the listener has to hold in memory before the answer arrives.

Models are unreliable on all four. Ask for A2 and you will often get B1 vocabulary delivered in short sentences, which is a different thing entirely. Read the script against the CEFR descriptors published by the Council of Europe and edit rather than trusting the label.

Three quick checks that catch most problems. Count the unfamiliar words in the first thirty seconds: more than three and the class will disengage before the task starts. Look at the longest turn: anything over about forty words at A2 is a monologue pretending to be a conversation. And check where the answers sit, because generated scripts love to put every answer in the final ten seconds.

If you want the levelling done more carefully, the same CEFR logic applies to written material, and our guide to adapting teaching materials with AI covers the checking routine in detail.

Accents, and why one voice is not enough

If every recording your class hears is the same polished accent, you are teaching them one accent. English is not one accent. Most text to speech systems offer several varieties, so rotate them across a term rather than settling on the one you find easiest to listen to.

Be careful about what you tell learners, though. A synthetic Irish or Indian or Nigerian voice is an approximation, and presenting it as a model of how people in a place speak is not accurate. Say plainly that it is a generated voice. For genuine variety, real recordings still win, and there is no substitute for hearing a person who is tired, in a hurry and standing next to a road.

Recording your own learners

Some of the best listening material in any classroom is the class. Learners interviewing each other, then transcribing what they said, produces material at exactly the right level about things they care about.

One rule though. A recording of a learner’s voice is personal data, and a recording of a child is sensitive. Get permission, keep it local where you can, and do not upload it to a tool your institution has not approved. The Irish Data Protection Commission guidance is short and worth reading once properly.

Where AI does not belong in a listening lesson

It does not belong in the part where learners fail. The purpose of a listening task is partly to let people not understand something and then work out how to cope, and a teacher who rescues that moment with a transcript on screen has removed the lesson.

It also does not belong in assessment without care. Generated audio is easier than authentic speech, so a learner who scores well on it has demonstrated something narrower than you might assume.

And it does not replace the second best listening resource in the building, which is you. Teacher talk, adjusted live to a class you are watching, does something no recording does.

Where to start this week

Take the topic your class is on now and build one dialogue. Six prompt elements, one target structure, two voices, roughly ninety seconds. Write three gist questions and five detail questions. Run it, then keep the transcript for a second lesson.

If it works, the rest follows quickly, and the whole lesson around it can be drafted the same way using our guide to ESL lesson plans with AI. If speaking practice is the next gap, our piece on using AI as a speaking partner covers the other half of the conversation, and the broader picture sits in our guide to AI in English language teaching.

Every free tool we run is on the AI tool overview page, and if you want the skills assessed rather than just practised, the AI-Skilled Teacher Certificate is the structured route.