Artificial intelligence is rapidly reshaping how language learning is delivered, assessed and experienced. Chatbots, automated writing evaluation tools, adaptive platforms and AI-generated feedback are increasingly embedded in mainstream English language education. Yet a body of emerging research suggests that the learners who might benefit most from these technologies – adult migrants and refugees accessing ESOL in community and FE settings – are at serious risk of being left behind.
A significant study published in 2026 in Language Teaching Research, the SAGE journal, examined AI literacy among adult EAL learners from migrant and refugee backgrounds. Its findings were striking: the majority of learners in this cohort had either never heard of generative AI, or had never used it. The authors argue that access to generative AI and the ability to use it are becoming crucial for learning, work and leisure — and that there is therefore a growing call to include AI literacy within language learning programmes. Without deliberate intervention, the digital divide risks becoming an AI divide.
This finding sits in uncomfortable tension with the broader narrative around AI as a democratising force in language learning. The technology is frequently promoted on the basis that it can deliver personalised, low-cost, always-available instruction to learners regardless of location or circumstance. In practice, the research suggests that for a significant proportion of ESOL learners, that promise remains unrealised, not due to any limitation in the technology itself, but because the learners in question have had no meaningful exposure to it.
A complementary picture emerges from a major scoping review published in April 2026 in Language Teaching Research, which analysed 272 empirical studies on AI in language education published between 2005 and 2024. It found a significant shift from rule-based systems to advanced AI applications over the two decades studied, with chatbots accounting for 44.1% of tools examined and a strong emphasis on productive skills — particularly writing and speaking. But the study’s demographic breakdown revealed a substantial blind spot: university students constituted 67.6% of all research participants, while adult learners — typically those in part-time or continuing education, and the core population of ESOL provision — accounted for just 4.8% of studies.
For ESOL providers, this underrepresentation in the research base matters. The tools currently being developed, validated and marketed are overwhelmingly designed for younger, academically-enrolled learners in relatively resource-rich environments. The evidence base for using these tools with adults who may have low prior digital literacy, interrupted formal education, or specific linguistic profiles associated with forced migration is thin.
There is also a teacher dimension. A study published in Frontiers in Psychology in June 2026, examining ESL teachers’ integration of AI tools in secondary schools in Ghana, found that teacher self-efficacy plays a critical mediating role in whether AI tools are used effectively in the classroom. Anecdotally, this is also very much the case in the ESOL sector. Teachers who lack confidence in using AI, or who have not been trained to do so, tend to use it superficially or not at all. This has direct implications for the ESOL workforce, which is already under significant pressure and where continuing professional development is often under-resourced.
The message for ESOL providers and policymakers is clear: the arrival of AI in language education is not automatically inclusive. If AI literacy is not embedded in ESOL curricula and teacher training, the sector risks widening the very inequalities it exists to address.
The Language of Change