News Round-up:
What You Missed During These Holidays
Cambridge added AI literacy to its Life Competencies Framework in July. Pearson began describing interaction with generative AI through formal English learning objectives. IELTS continued its retreat from paper-based testing, while new studies examined what AI is actually doing in EFL classrooms and what teachers make of it.
Quite a lot happened while schools were closed.
For anyone who sensibly spent part of the summer ignoring education news, these are the developments worth catching up on before the new academic year gets properly under way.
AI literacy enters the language-learning conversation
One of the more consequential announcements came from Cambridge English. Its updated Life Competencies Framework now includes AI literacy, with attention given to understanding AI and identifying fake content.
The change may look modest, but it places AI in a different part of the educational equation. Until recently, most discussion in ELT concerned what teachers could do with AI: generate materials, prepare lessons, create exercises or provide feedback. Cambridge is now addressing what learners themselves need to understand about the technology.
Pearson has approached the issue from another direction. Through the Global Scale of English, it has introduced learning objectives concerned specifically with interactions with generative AI.
Giving instructions to an AI system, assessing the response, clarifying what you want and reformulating an unsuccessful request all involve language. When those exchanges take place in English, they can also become communicative tasks.
That opens an interesting curricular question. AI literacy may no longer sit somewhere outside ELT, filed under digital skills. Some of it could become part of what learners are expected to do through English.
Pearson puts AI inside its ELT products
Pearson has also been building AI directly into its English-language learning environment. Its 2026 offering includes an AI Smart Lesson Generator for teachers and an AI Speaking Tutor providing personalised speaking practice and feedback to learners.
There is an important distinction here. Teachers have been using general-purpose systems such as ChatGPT to produce classroom materials for some time. Publisher-integrated AI puts the technology inside the same ecosystem as courses, assessment and teacher resources.
For an industry built for decades around the coursebook and its supporting materials, that could alter the relationship between publisher and classroom. A publisher can potentially provide the course, generate elements of the lesson, give the learner additional practice and participate in feedback and assessment.
The coursebook does not disappear. But it may no longer be the centre of the relationship.
IELTS moves further away from paper
For examination teachers, one of the more immediate developments concerns IELTS.
The test is moving towards computer delivery, with the transition away from traditional paper-based testing expected to be completed during 2026. A handwriting option for the Writing component is being retained selectively during the transition.
For candidates, changing the delivery format affects preparation. Reading extended texts on a screen feels different from working on paper. Candidates also need to type accurately under pressure, move confidently around the testing interface and review written work electronically.
Those are not English-language skills in themselves, yet poor digital familiarity can interfere with a candidate’s ability to demonstrate the English they have.
IELTS is part of a wider direction in international language assessment. Computer delivery is steadily becoming the normal testing environment rather than an alternative version of the exam.
TOEFL reports uptake after its 2026 changes
TOEFL had its own changes to digest over the summer.
In June, ETS reported on international adoption of the enhanced TOEFL iBT introduced in January 2026. The organisation has presented the revised test around a more modern testing experience and greater flexibility.
Claims about adoption come from ETS itself, so they should be read as provider-reported figures rather than independent evidence of market share.
Even so, the direction of competition between the major English examinations is becoming easier to see. Recognition remains essential, but candidates also expect convenient delivery and faster results. Exam providers are responding accordingly.
For preparation centres, test-format changes are therefore about more than keeping a syllabus up to date. The testing experience itself has become part of what providers compete on.
Digital assessment reaches younger learners
The shift towards digital testing is also moving down the age range.
Cambridge English announced in June that its Digital for Young Learners qualifications had been shortlisted in three categories at the International e-Assessment Awards.
The nominations are one detail. The development of digital assessment for children is the more interesting story.
Computer-based language examinations were once associated mainly with adults, university applicants and professional candidates. Younger learners are increasingly being assessed in similar environments.
That brings a complication. When children take language tests digitally, exam designers and teachers need to distinguish between English proficiency and confidence with the technology used to measure it. A learner should not perform worse in an English examination simply because the interface is unfamiliar.
As digital assessment reaches younger candidates, that problem deserves more attention.
What do 103 studies tell us about AI in EFL?
The summer brought something the ELT debate on AI has often lacked: a larger body of evidence to examine.
A systematic review published in Computers and Education: Artificial Intelligence looked at 103 studies dealing with artificial intelligence in EFL learning.
The number itself shows how quickly the field has grown. There is already enough published work to step beyond individual classroom experiments and start asking what patterns emerge across the research.
There is, however, an important limitation. Much of the research has been conducted in higher education. The evidence involving school-age learners is thinner.
That matters because schools are already dealing with generative AI. Children and teenagers use it outside school, teachers are experimenting with it in lessons and companies are developing products for younger learners. Yet evidence drawn from university students cannot automatically tell us what works with a 10-year-old or a 15-year-old.
Adoption, in other words, may be moving faster than the research base for some of the learners being asked to use the technology.
Source: Computers and Education: Artificial Intelligence, systematic review of 103 studies on AI in EFL learning, 2026.
Teachers don’t need to become prompt engineers
A related piece of research published in Frontiers in Education examined what it calls pedagogical prompting.
The distinction is useful. Much of the early training around generative AI taught people to construct increasingly elaborate prompts, sometimes giving the impression that teachers needed to become amateur prompt engineers to use the technology effectively.
For classroom purposes, technical sophistication is a poor measure of success. An elaborate prompt that produces an impressive worksheet is still a poor prompt if the worksheet serves no worthwhile learning objective.
Pedagogical prompting puts the teacher’s judgement back at the centre: knowing what learners need to practise, recognising weak or inaccurate output, adjusting material to level and deciding whether AI is appropriate for the task in the first place.
Those decisions were part of good teaching long before ChatGPT arrived.
AI can produce the task. The teacher still has to teach it
Cambridge researchers considered a similar problem in work on generative-AI-supported task-based language teaching.
AI can produce dialogues, texts, comprehension questions, vocabulary activities, role-play scenarios and feedback in seconds. The amount of material a teacher can generate is effectively unlimited.
Teachers choose tasks for particular learners. They notice when an activity is failing, change pace, draw a quiet student into the conversation, recognise when apparent understanding is superficial and decide what needs to happen next. Much of teaching takes place in those small judgements rather than in the worksheet or slide on the screen.
AI makes producing classroom content easier. It does not remove the need to know what to do with it.
Assessment has a new authenticity problem
Perhaps the most difficult questions are appearing in assessment.
Pearson has been examining formative assessment in the age of generative AI, where a familiar piece of evidence, the finished student text, has become harder to interpret.
A student submits an excellent essay. The teacher can see the quality of the English, but may know much less about how that English was produced.
That weakens an assumption on which much writing assessment has traditionally depended: that the language in the final submission provides reasonable evidence of the learner’s current ability.
The response may be surprisingly traditional. Drafts become more informative. So do classroom writing, conversations about language choices, revision decisions and a student’s ability to explain or reproduce what appears in the submitted work.
Generative AI can produce increasingly convincing final texts. That may make the process behind the text more valuable as evidence.
