AIDI, a cefr-based chatbot for Italian as a second language: Pilot survey on learners' perceptions

Learning Italian as a second language remains challenging, especially in achieving authentic communication and accessing personalised support. We present a pilot study conducted with AIDI (Artificial Intelligence for Italian Dialogue), a CEFR-aligned chatbot (A1–B2) that provides interactive practice through realistic scenarios. To ensure level-appropriate, consistent output, we employed prompt engineering, i.e., CEFR-calibrated scenario templates with role assignment, levelgraded lexis, and response constraints to maintain register and mitigate bias. In a pilot study with 31 L2/LS learners, participants engaged with AIDI via the institutional LMS and completed a TAMinformed questionnaire assessing perceived usability and linguistic usefulness. Learners reported high engagement and positive attitudes toward AIDI as a conversational partner, as well as perceived support for autonomous practice and perceived gains in expressive fluency. These findings suggest that CEFR-aligned chatbots can support scalable, personalised practice; however, larger longitudinal studies incorporating learner-production data are required to establish sustained pedagogical effects.