There is a version of this story where I tell you that I opened a language model, typed a brief, and the handouts appeared fully formed, ready for Monday’s lesson. That version is fiction.
What actually happens is considerably more interesting, and considerably more labour-intensive.
I use AI to create the ESLTutorHub materials. The process looks nothing like the single-click content generation that tends to get discussed whenever AI and education appear in the same sentence.
The materials start with a subject: something adults actually want to argue about. The kind of topic that makes a student forget, briefly, that they are supposed to be learning English. From there, the text has to earn its place: the vocabulary has to be intentional, the ideas have to be worth responding to, and the activities have to grow organically from the language on the page rather than being bolted on afterwards. A synonym exercise that has no relationship to the text it accompanies is just busywork. The compound word task has to open a door into the same semantic territory the student has just walked through.
This is where the language model and I tend to disagree most productively. Left to itself, a model will produce something competent, occasionally something impressive, and regularly something that a linguist has to correct, redirect, or simply discard. The gap between what an LLM generates and what a teacher can actually use in a room with a real adult is not a small gap. It is the gap that my work exists to close.
I find this genuinely encouraging rather than threatening. The model handles the mechanical scaffolding with a speed and patience I cannot match. I handle the judgment: whether a particular collocation is attested, whether a discussion question is genuinely uncomfortable or merely sounds like it is, whether the B1 version of a text has been simplified in a way that preserves the idea or merely flattened it. That division of labour suits me.
The result, when it works, is a handout where the scaffolding is already built. The teacher still needs to know the text and the answers: that is not preparation that can be skipped, and any teacher who thinks otherwise will feel it in the room. What does not need to happen is the hours of construction that precede a lesson built from scratch.
Whether that justifies the hours it takes is a question I ask myself periodically. The answer, so far, has been yes. You can see the results at esltutorhub.com.