Teaching AI as a Language Skill

Generative AI tools are easiest for people who are already tech-savvy and fluent - which means the learners who could benefit most are often the ones locked out. This study explores a different premise: that conversing with an AI agent is itself a communicative task that can be taught explicitly, with scaffolding, inside a language classroom.

Working one-on-one with adult newcomer learners, I introduced AI tools as part of real goals - finding suitable jobs, updating a resume, exploring career pathways. Rather than handing over polished prompts, I taught small, transferable formulas (give your background, ask for “a real first step,” request “simple English”) that learners could reuse and adapt.

Over time, learners moved from never having heard of these tools to generating, refining, and fact-checking their own prompts - including pushing back when the AI got something wrong about their lives.

The barrier, it turns out, isn’t proficiency alone - it’s instructional access. When learners are given clear scaffolding, even those with lower language proficiency can use AI quickly and meaningfully, for purposes that matter to them. This work argues for treating AI literacy as a teachable language skill, and for including refugee and newcomer learners in that instruction rather than assuming they “aren’t ready.”