ChatGPT for A1–A2 Spanish: Mediation strategies and learner autonomy beyond the chat

What is it about?

Many university students now use ChatGPT for language learning, but it is not always clear how it helps with real communication skills. This study looks at multilingual university students learning Spanish at beginner levels (A1–A2) and asks how ChatGPT can support mediation, that is, helping to understand, explain, and relay meaning in a conversation. Students completed structured tasks in Spanish with ChatGPT, such as summarizing short texts, asking about vocabulary, and taking part in simple debates. Spanish was not the students’ first language, and it was also not the main language the AI was trained on, which made the setting more realistic for multilingual learning. We analysed student–ChatGPT conversations using CEFR mediation descriptors and also examined students’ written reflections (in their first languages). During the tasks, students often depended on ChatGPT for language support (words, phrasing, corrections). However, their reflections showed more advanced learning behaviours: setting goals, noticing problems in AI feedback, adjusting prompts, and comparing ChatGPT with other tools. Together, the results suggest that pairing AI interaction with structured reflection can reveal and strengthen learner autonomy.

Featured Image

Photo by Steve Johnson on Unsplash

Photo by Steve Johnson on Unsplash

Why is it important?

This work shows that even beginner language learners can use ChatGPT for meaningful communication tasks, not only practicing grammar and vocabulary, but also developing CEFR mediation skills such as summarizing, clarifying, and negotiating meaning. A key contribution is the contrast between what learners do during AI-supported tasks (often relying on the tool) and what they understand strategically when reflecting afterwards (goal-setting, evaluation, and tool choice). That matters for teaching, because it suggests that autonomy may be “hidden” during the interaction and becomes visible, and teachable, through reflection. The study also highlights a practical implication for CEFR-based pedagogy: existing mediation descriptors were designed for human-to-human interaction and may need adaptation to capture the dynamics of human–AI dialogue, including prompt management and digital agency. For educators, the findings support designing AI activities with explicit reflection steps so students learn not just with AI, but also how to use it critically and responsibly.

Perspectives