By Gemma De La Cruz-Duffy (Manchester Adult Education) and Rachel Öner (NATECLA)
How can teachers use AI to support English language learning?
Artificial intelligence (AI) is rapidly becoming part of the educational landscape, but for many teachers the challenge is not accessing AI tools or getting recommendations about the best tool for their needs, the challenge is knowing how to use them effectively. At the Bell Foundation 2026 conference, we ran a session exploring how AI can support inclusive EAL and ESOL teaching through two simple but powerful ideas: better prompts leading to better outcomes and the use of Gems or Agents.
Teachers do not need to become AI experts to benefit from these tools. Small changes in prompting can make a significant difference to the quality of the resources and interactions AI generates. Whether using ChatGPT, Gemini, Copilot, or another large language model, the quality of the response depends largely on the quality of the instructions we provide. As language teachers, this should feel familiar. We routinely adapt materials, scaffold learning, differentiate tasks, and consider learners' linguistic and cultural backgrounds. Effective AI prompting simply extends these existing teaching skills into a new digital environment.
The key message is to start small and stay purposeful. A useful starting point is to move beyond short, generic instructions and instead provide clear context about learners, language levels, and lesson objectives.
AI works best when guided by teacher expertise. Effective prompting is not about replacing professional judgement; it is about amplifying it. The more clearly we communicate our pedagogical intentions, the more useful AI becomes as a tool for supporting language learning.
Better prompts with the CLEAR framework
One of the most common reasons teachers become frustrated with AI is that the output feels generic, inaccurate, or pitched at the wrong level. Often, the issue is not the tool itself but the prompt that was used. A vague instruction such as “Explain the metaphors in Jekyll and Hyde for an ESL student” leaves too much for the AI to guess. What level is the learner? What language should be used? How much support is needed?
To address this challenge, one solution is to use the CLEAR framework, a simple structure that helps teachers design prompts that produce more useful and inclusive responses. So, what does CLEAR stand for?
Provide information about the learners, setting, and purpose of the task. For example, specify whether the learners are adults or teenagers, studying in the UK, preparing for GCSEs, or developing English for work and everyday life.
Clearly identify learners' language proficiency. This might include The Bell Foundation’s five proficiency bands, CEFR levels, ESOL Entry levels, or a description of what learners can and cannot yet do. The more specific the information, the more accurately the AI tool can adapt its language.
Teachers can request British English, particular tenses, sentence structures, vocabulary limits or text length. This helps ensure that materials remain accessible and aligned with learning objectives.
Just as we ‘differentiate’ classroom activities, prompts can request scaffolds, extension tasks, vocabulary support, or alternative versions for learners with different literacy levels. AI becomes far more useful when it is asked to build in support from the outset.
Inclusive teaching matters. Prompts can encourage diverse names, cultures, family structures, and experiences, while also ensuring materials are accessible and sensitive to learners' backgrounds. However, do bear in mind that AI tools have been trained predominantly on western languages and cultures, which has resulted in concerns about bias. AI results must be assessed critically when being used to help create resources that reflect the diversity of EAL and ESOL classrooms.
The CLEAR framework transforms prompting from a technical skill into a pedagogical one. Rather than simply asking AI to generate content, teachers guide it using the same principles that underpin effective lesson planning. The result is more relevant, more inclusive, and more usable classroom resources that require less editing and adaptation afterwards.
Role-based prompting: moving from content to conversation
While many teachers use AI to generate texts, worksheets, or quiz questions, one of its most powerful applications is as a conversational partner. By assigning AI a role, teachers can create realistic scenarios that allow learners to practise language in authentic contexts. For example, AI can act as a landlord, shop assistant, employer or workplace supervisor, giving learners opportunities to rehearse interactions they may encounter outside the classroom.
Roleplay has long been recognised as an effective teaching strategy in language education. It offers learners a meaningful reason to communicate, encouraging them to use language for real purposes rather than simply completing discrete grammar exercises. Research in language acquisition helps explain why roleplay is so effective.
Krashen's theory of the affective filter suggests that learners acquire language more readily when anxiety is reduced. Practising with AI allows learners to make mistakes, restart conversations, and experiment with language without fear of embarrassment – especially for newly arrived students who might have to participate in a practical science lesson for the first time, for example.
Similarly, Long's Interaction Hypothesis emphasises that language develops through interaction and negotiation of meaning, while Vygotsky's concept of scaffolding highlights the importance of support that helps learners move beyond what they can currently do independently. AI roleplay has the potential to support all of these processes.
For ESOL and EAL learners, role-based prompting can also extend speaking practice beyond the classroom. Learners can rehearse booking appointments, attending college or job interviews, reporting illness or workplace issues, or engaging in everyday conversations as many times as they need. The interaction is immediate, personalised, and available whenever learners are ready to practise.
Used thoughtfully, role-based prompting moves AI beyond content generation and transforms it into a tool for meaningful communication, confidence-building and language development.
Creating reusable AI tutors with Gems and Agents
Once teachers have developed effective prompts, the next step is creating reusable AI tutors. In Gemini these are known as Gems and in Copilot they are known as Agents. Rather than rewriting the same instructions every time, teachers can save a set of rules and processes that the AI will follow consistently.
A useful way to think about a Gem or Agent is as a custom tutor that remembers its role and instructions. For example, an ESOL tutor might create a Gem that automatically adapts language to Entry 2 level, asks one question at a time, provides supportive feedback, and adjusts the difficulty of tasks according to learner responses. Once the setup is complete, learners can engage in a range of roleplays and conversations without the teacher having to repeatedly provide detailed prompting instructions.
The benefits are practical as well as pedagogical. Gems and Agents can save preparation time, ensure consistency, and provide learners with additional opportunities for independent practice. They can also be shared with learners, creating a personalised support tool that extends learning beyond the classroom. Ultimately, the effectiveness of these tools depends not on the technology itself but on the expertise that teachers bring when designing them.
The views and ideas presented in this blog are the author's own and do not necessarily represent the views of the Foundation.