AN INDUSTRY REPORT FROM THE TEFL INSTITUTE GROUP · JULY 2026
The full text of our industry report is published below, free and in full. It pulls together published data from UNESCO, the British Council, Gallup, PwC, RAND, Cambridge and the leading market analysts into one honest picture of where AI is taking English language teaching and TEFL careers. Every figure is attributed to the organisation that published it.
Download the PDF version
10 pages · fully sourced · free, no email required
Artificial intelligence has moved from the edges of English language teaching to its centre in under three years. Gallup found that 60% of United States public school teachers used AI tools during the 2024-25 school year, and that weekly users save an average of 5.9 hours per week, the equivalent of about six working weeks per year. The British Council’s survey of 1,348 teachers across 118 countries found English language teachers already using AI for materials creation (57%), learner practice (53%) and lesson planning (43%).
Yet the same evidence shows this is an augmentation story, not a replacement story. In that British Council survey, 60% of English teachers worldwide reject the idea that AI translation makes language learning unnecessary, and 51% are sceptical that AI will be able to teach English without a human teacher even by 2035. The British Council’s Future of English research programme concludes that teachers will remain central to English learning even as automation increases. Our own State of TEFL 2026 research reaches the same verdict: AI saves TEFL teachers an estimated 3 to 5 hours per week and augments rather than replaces human teachers.
The commercial stakes are significant. Analyst forecasts from MarketsandMarkets, Mordor Intelligence, Grand View Research and Global Industry Analysts place the AI-in-education market at roughly $2 to 8 billion today, growing to between $6 and 57 billion by 2030-33 depending on scope. The Business Research Company expects the global English language training market to grow from about $89 billion in 2025 to $123 billion by 2030. The biggest risk the sector faces is not job losses but a training gap: the British Council found only 20% of English language teachers feel sufficiently trained to use AI in their teaching.
| Metric | Figure | Source (year) |
|---|---|---|
| Global English language training market | $89.2B (2025) growing to $122.7B by 2030 | The Business Research Company (2025) |
| AI-in-education market forecasts | $2.2B to $8.3B today; $5.8B to $57.2B by 2030-33 | MarketsandMarkets, Global Industry Analysts, Mordor, Grand View (2025-26) |
| US teachers who used AI in the last school year | 60% (32% weekly or more) | Gallup (2025) |
| Time saved by weekly AI users | 5.9 hours per week | Gallup / Walton Family Foundation (2025) |
| Wage premium for AI-skilled workers | 56% (up from 25% the prior year) | PwC Global AI Jobs Barometer (2025) |
| Job growth in the most automatable roles | +38% (2019-2024) | PwC (2025) |
| English learners worldwide | 2 billion+, taught by about 12 million teachers | State of TEFL (2026) |
| Teachers sceptical AI can teach English without a human by 2035 | 51% | British Council, 118-country survey (2025) |
| English teachers who feel sufficiently trained on AI | only 20% | British Council (2025) |
Analyst houses disagree on the exact size of the AI-in-education market because they define it differently, but they agree emphatically on direction. Estimates for the current market range from $2.2 billion to $8.3 billion, with forecasts of $5.8 billion to $57.2 billion by 2030-33 and compound annual growth rates of 17.5% to 42.8%, according to MarketsandMarkets, Mordor Intelligence, Grand View Research and Global Industry Analysts.
The Business Research Company sized the global English language training market at $89.2 billion in 2025, forecast to reach $122.7 billion by 2030. The digital slice is growing far faster: Technavio forecasts digital English language learning to grow by $39.5 billion between 2025 and 2029, a 24.5% annual rate it explicitly attributes to AI redefining the market. Language learning apps alone are projected by Business of Apps to grow from $5.4 billion in 2024 to $14.1 billion by 2030.
Duolingo reported 2025 revenue of $1.04 billion, up 39% year on year, with 52.7 million daily active users and 12.2 million paying subscribers. English remains its largest studied language, with 48.6 million learners. Yet its market value fell from a $21.7 billion peak in 2025 to $5.7 billion by April 2026, driven by investor anxiety about AI disruption even as usage and revenue grew. The lesson for the ELT sector: AI narratives can move sentiment faster than they move classrooms. HolonIQ put EdTech venture investment at $2.6 billion in 2025, with the biggest bets concentrated in AI and workforce training.
The single most useful dataset on AI in English language teaching is the British Council’s global survey of 1,348 English language teachers across 118 countries. It reveals a profession that is pragmatic about the technology: open to AI as a support tool, confident in the human role, and under-trained for the transition.
| Finding, British Council 118-country survey (2025) | Share of teachers |
|---|---|
| Use AI for materials creation | 57% |
| Use AI for learner practice | 53% |
| Use AI for lesson planning | 43% |
| Reject that translation removes the need to learn English | 60% |
| Sceptical AI can teach English without a human by 2035 | 51% |
| Feel sufficiently trained to use AI | 20% |
| Feel inadequately trained to use AI | 54% |
In the United States, teacher AI use jumped from 25% in autumn 2023 to 60% in the 2024-25 school year, with 32% using it weekly or more, according to Gallup and RAND. RAND also found district-provided AI training doubled in a single year, from 23% to 48%, though wealthier districts train far more of their teachers (67%) than high-poverty districts (39%). Use is strongly stratified by level: RAND data reported by K-12 Dive shows 69% of United States high school teachers using generative AI against 42% of elementary teachers.
In the UK, British Council research found 79% of secondary teachers have rethought how they set assignments because of AI, with 59% integrating it constructively and 38% designing deliberately AI-proof tasks. Primary language classrooms remain largely untouched: the British Council’s Language Trends England 2025 found 75% of England’s primary schools never use AI in language teaching. The UK Department for Education’s Technology in Schools Survey found 21% of primary and 25% of secondary schools already offer staff generative AI training, with roughly half planning to introduce it.
The most common teacher use cases are consistent across surveys: preparing to teach and creating materials lead everywhere, followed by differentiation and lesson planning. Across nine measured tasks, the Walton Family Foundation found 60-84% of United States teachers report time savings from AI, while only 7% say it slows them down.
AI speech systems now analyse pronunciation accuracy, fluency and speech rate in real time, and research published in Frontiers in Psychology links that instant feedback loop to improved learner motivation and engagement. Comparative analyses conclude that AI tutors can sustain long spoken conversations and correct pronunciation at very low cost, but lack a coherent model of the learner’s long-term progress and cannot replicate a human teacher’s cultural fluency or accountability, which supports a hybrid model rather than substitution. Researchers writing in Behaviour and Information Technology describe the emerging pattern as hybrid teaching intelligence: teachers delegate marking and personalisation to AI while retaining judgement, but report clear training gaps.
Purpose-built planning tools have moved this from novelty to workflow. Our own TEFL.ai platform generates CEFR-aligned (A1-C2) lesson plans, worksheets, role-plays and differentiated materials from a single dashboard, alongside exam-support and career tools, and is explicitly positioned as a support tool that will not replace great teachers. The State of TEFL 2026 report finds TEFL teachers already using AI across lesson planning, pronunciation feedback, assessment, adaptive learning and content creation.
Cambridge University Press and Assessment has published 12 guiding principles for AI-powered automarking of language tests, explicitly advocating a human-in-the-loop hybrid rather than fully autonomous scoring. The Duolingo English Test is already a fully AI-driven assessment, with human-AI scoring agreement above 0.85 and acceptance by more than 5,500 institutions. ETS has used its e-rater engine to score essays alongside human raters for years. But generic chatbots are not examiners: a 2025 study in Innovations in Education and Teaching International found ChatGPT consistently assigned higher grades than human raters and failed to reliably distinguish high- from low-proficiency writing. Note that IELTS has not publicly adopted fully automated AI scoring for its core exam, and claims to the contrary should be treated with caution.
UNESCO argues that generative AI forces a rethink of what is worth measuring, pushing assessment toward authentic, process-based evaluation. Analysts note that AI now makes continuous, contextual assessment through portfolios and adaptive tasks a credible alternative to one-off high-stakes tests.
The EF English Proficiency Index 2025, based on 2.2 million adult test-takers in 123 countries, finds speaking is the weakest skill in over half of surveyed countries and that youth proficiency is declining rather than rebounding after the pandemic. Major TEFL destination markets including Japan, Mexico, Saudi Arabia and Thailand sit in the lowest proficiency band and declined year on year, which sustains demand for qualified English teachers. UNESCO estimates the world needs more than 40 million additional teachers by 2030, and the OECD calls teacher shortages an urgent and growing challenge across member countries.
Our State of TEFL 2026 research sizes the global ELT market at roughly $95 billion in 2026, growing to $181 billion by 2034, with more than 2 billion English learners worldwide, 12 million English teachers, and over 2 million TEFL positions opening globally each year. Regional demand is growing fastest in Asia-Pacific (5.8-8% CAGR) and the Middle East and Africa (7-9% CAGR).
PwC’s 2025 Global AI Jobs Barometer, built on roughly one billion job advertisements, found that jobs grew even in the most automatable roles, up 38% from 2019 to 2024, while AI-skilled workers commanded a 56% wage premium in 2024, up from 25% a year earlier. PwC also found degree requirements declining fastest in AI-exposed postings, suggesting employers increasingly value demonstrated AI fluency over credentials, which plays to the strengths of practical TEFL certification.
Inside the classroom, the time dividend is the headline: 3 to 5 hours saved per week according to our own State of TEFL 2026 data, and 5.9 hours for weekly users in Gallup’s United States survey. That matters because teaching is a profession under strain. Our State of TEFL research found 77% of teachers report frequent stress, 88% find the job overwhelming at times, and 93% report inadequate institutional support, even though 83% feel a strong sense of purpose. Framed properly, AI adoption in TEFL is as much a wellbeing intervention as an efficiency play. One industry report claims 7 in 10 new English teachers now co-teach with AI. We flag this as an emerging, anecdotal claim rather than a rigorously sampled statistic.
Generative AI produces plausible but sometimes false content, and researchers note teachers often lack formal training to reliably detect hallucinations in AI-generated learning materials. Verification habits, not blind trust, are the core new skill.
AI detectors are not a safe foundation for high-stakes decisions. Reporting in Nature puts GPTZero’s false-positive rate on human-written text at an estimated 16%, and a Stanford-linked study found common detectors falsely flagged over 61% of TOEFL essays written by non-native English speakers as AI-generated, a severe equity risk for exactly the learners TEFL serves. We note that we have seen the Stanford figure via secondary reporting rather than the original paper. Research in Humanities and Social Sciences Communications shows AI-assisted human review can raise detection accuracy of machine-translated text from about 52% to about 75%, but the deeper answer is assessment redesign, not detection arms races.
A UN Special Rapporteur report on AI in education highlights risks from EdTech data collection, web-scraping and surveillance of children’s data, and a child-rights audit of generative AI EdTech tools by the 5Rights Foundation found widespread opaque data practices and commercial exploitation of children’s data. UNESCO recommends a minimum age of 13 for independent generative AI use in education.
Research in Frontiers in Computer Science finds AI education tools are predominantly built for English and other major languages, structurally disadvantaging multilingual learners, and the OECD flags access and bias risks. The counter-evidence is encouraging: when culturally adapted, adaptive learning platforms show effect sizes of d=0.40 to 0.85 for reducing educational inequities. Design choices decide whether AI narrows or widens gaps.
The British Council found only 20% of English language teachers feel sufficiently trained to use AI, while 54% feel inadequately trained. RAND found training access unequal by district wealth, and UNESCO has urged governments to regulate generative AI in schools precisely because rollout is outpacing teacher readiness. For TEFL training providers, closing this gap is both a duty and the clearest commercial opportunity of the decade.
UNESCO published the first global policy guidance on generative AI in education in 2023, identifying eight controversies from digital poverty and regulatory lag to model opacity and deepfakes, and recommending a minimum age of 13 for independent use. Its 2024 AI Competency Frameworks for teachers and students define progressive skill levels to guide training and curriculum design, and its assessment think-piece reframes the integrity debate around what is worth measuring in an AI age.
The Future of English: Global Perspectives programme, built on roundtables with 92 policymakers and experts from 49 countries, concluded that English will retain its position as the world’s most widely spoken language over the next decade, and found a strong connection between the desire to learn English and the need for teachers, even when new technologies are considered. It also warns that technology risks widening the divide between those with and without access. The British Council’s 118-country teacher survey supplies the adoption and attitude data used throughout this report.
Cambridge University Press and Assessment’s 12 guiding principles for AI-powered automarking are the clearest institutional statement from a major English-testing body: AI increases marking speed and consistency, but humans must remain in the loop for high-stakes judgement.
ETS’s e-rater engine has applied AI and natural language processing to essay scoring for years, evaluating grammar, mechanics, style and organisation alongside human raters. That is proof automated scoring is already embedded in mainstream high-stakes English testing rather than a speculative future.
The convergence across all four institutions is striking: none predicts teacher replacement, all prioritise teacher training and governance, and all treat AI as infrastructure to be shaped rather than a wave to be survived.
The weight of evidence, from our own State of TEFL data on hours saved and augmentation, to Gallup’s AI dividend and PwC’s finding that jobs grew even in highly automatable roles, points one way. By the late 2020s, using AI as a planning, assessment and practice co-pilot will be standard professional practice in TEFL rather than a differentiator.
Teachers are already delegating marking and personalisation to AI while retaining judgement-based work. The core value of a TEFL-certified teacher shifts toward structuring practice, culturally nuanced feedback and motivational accountability, precisely the tasks AI handles poorly.
Cambridge’s human-in-the-loop principles, the Duolingo English Test’s validated AI-driven model and UNESCO’s what-is-worth-measuring agenda all point toward continuous, portfolio-style assessment displacing snapshot testing over the decade.
AI-skilled workers already command a 56% cross-industry wage premium on PwC’s figures, and degree requirements are falling fastest in AI-exposed roles. With only 20% of English teachers feeling adequately trained, scarcity should hand AI-literate TEFL teachers a hiring and pay advantage. We flag this one as inferential: TEFL-specific wage data does not yet exist.
Meaning lives in tone, register and trust, not just literal words, and commentators consistently find translation earbuds cannot substitute for genuine language ability. Sixty per cent of teachers in the British Council survey agree translation will not make learning unnecessary, and English learning demand keeps growing on every market forecast in section 1.
Every serious dataset in this report tells the same story from a different angle. Money is flowing into AI-powered English learning at venture and enterprise scale. Teachers are adopting AI faster than their institutions can train them. The technology is genuinely good at the repetitive layer of teaching work, planning, materials, drilling and first-pass marking, and genuinely poor at the human layer: cultural nuance, motivation, accountability and judgement.
For anyone considering a TEFL career, the rational response to AI is not hesitation but acceleration. Demand for English remains enormous, with over 2 billion learners and 2 million positions opening each year, and the teachers who thrive will be those who treat AI as part of their professional toolkit from day one. For schools and training providers, the mandate is equally clear: close the training gap before it becomes the sector’s defining inequality.
This report synthesises published research from UNESCO, the British Council, Cambridge University Press and Assessment, ETS, the OECD, RAND, Gallup, the Walton Family Foundation, PwC, EF Education First, the UK Department for Education, Nature, Frontiers, Taylor and Francis, the 5Rights Foundation and leading market analysts including The Business Research Company, MarketsandMarkets, Mordor Intelligence, Grand View Research, Global Industry Analysts, Technavio, HolonIQ and Business of Apps, alongside The TEFL Institute Group’s own State of TEFL 2026 industry research.
Market forecasts are presented as ranges because analyst definitions differ. Two claims are flagged in the text as lower-confidence: the 7-in-10 co-teaching figure, and the Stanford detector study, which we have seen through secondary reporting rather than the original paper. IELTS has not publicly confirmed fully automated AI scoring for its core exam, and this report makes no such claim. Each organisation named above publishes the underlying research on its own website, and readers are encouraged to consult the primary source directly for any figure they intend to rely on.
The TEFL Institute Group (2026). The Future of AI in English Language Teaching and TEFL. Published July 2026. Available at: https://tefl.ai/future-of-ai-report/
Journalists, researchers and educators are welcome to quote any figure in this report with attribution. For press enquiries or the full source list with references, contact us through tefl.ai.
The TEFL Institute Group operates The TEFL Institute of Ireland, TEFL.ie and tefl.ai, delivering accredited TEFL and TESOL certification, teach-abroad programmes and free AI-powered teaching tools to a global community of English language teachers. This report is also published by our sister brands: The TEFL Institute · TEFL Institute of Ireland.
Put the findings into practice with our free tools: the AI lesson plan generator, the full AI tool suite and the TEFL career roadmap.
