Creating custom GPTs to support teaching and learning

As a language practitioner, there are countless ways in which tools like ChatGPT, NotebookLM, and ElevenLabs have supported my work. GenAI is really strong at language, hence very useful for resource creation (e.g. tailored interactive vocab apps, audio and text), and for rote tasks such as structuring course programmes and mapping learning objectives to specific templates.

It also acts as a reflective partner in ideation, and helps me test and sharpen task designs. Before using GenAI, all of this was done manually, and many routine tasks were rather time consuming.

I have developed two custom GPTs: one for supporting my own task design and that of colleagues; and one I shared with students.

Custom GPT to support task design

I created a custom GPT that acts as an instructional design coach grounded in the CEFR action-oriented approach, which views learners as social agents performing meaningful communicative tasks. Its purpose is not to generate ready-made classroom activities, but to support me in refining and stress-testing task ideas. It helps ensure that language learning is conceptualised as social action with a clear communicative purpose and that task design remains aligned with the intended learning objectives. I use this custom GPT as a thinking partner to flesh out ideas and identify gaps in my design (e.g. limited learner agency, unclear communicative purpose, missing mediation or plurilingual component).

The GPT is configured to respond socratically, i.e. asking focused questions before suggesting revisions, so that the intellectual and pedagogical decisions remain with me.

I developed an equivalent version in Gemini as "Gem" to be able to compare behaviours and outcomes.
Since custom GPTs come with many strings attached, I have run the same custom instructions as a project, which is better for cross-chat memory but unlike custom GPTs they cannot be shared directly with others.

Custom GPT to support students

I developed a custom GPT translation coach for my Year 2 and Year 4 German students that supports reflective revision rather than simply correcting their work. It draws on a shared grammar reference, model answers and common error patterns, and responds by asking targeted questions, highlighting key linguistic issues, pointing students to the relevant grammar sections, and suggesting follow-up exercises for focused practice. The aim is to strengthen students’ self-editing skills, grammatical awareness and to provide additional practice opportunities. Students valued the additional practice and the opportunity to build AI literacy, but concerns were raised about using limited tutorial time for this exploration. We also struggled with a key question: how can you know whether the AI is correct if you don’t yet know the answer yourself? That uncertainty about when to trust the output led to productive conversations about verification strategies and the continued importance of tutor feedback.

Without AI I could not have experimented with as many creative and fun ideas in my teaching. While some routine tasks like drafting objectives, generating vocabulary lists and activities are much quicker, my overall workload hasn’t really reduced. The time saved is usually reinvested in exploring, testing, and trying to stay current in a fast-evolving field.

I am able to spend more time thinking about pedagogy rather than formatting or drafting from scratch.  I’ve learned practical prompting strategies and how to get better outputs, and the process has strengthened my confidence in the value of my own expertise and judgement.

If you are interested in implementing AI tools into workflows or the classroom, I would recommend involving students or colleagues early. Their scepticism will teach you as much as the tool does.