AI-Chatbot Professional-Development Recommender — University College of Teacher Education Burgenland

Researchers Thomas Leitgeb and Michael Leitgeb at the Centre for Digital Education, University College of Teacher Education Burgenland (Austria), built and evaluated an AI-based chatbot that gives teachers context-sensitive, personalised recommendations for professional-development (PD) courses, grounded in TPACK theory and self-determination theory.

The premise: generic, one-size-fits-all PD catalogues get low uptake because they ignore a teacher's subject, career stage and specific pedagogical needs.

Over a six-week field deployment (15 April-31 May 2024), 1,125 of 4,235 eligible teachers used the chatbot, generating 2,030 valid interactions (an average of 1.69 queries per teacher), analysed using a convergent parallel mixed-methods design -- logistic regression, correlation analysis, and qualitative content and sentiment coding.

The researchers measured 85.6% accurate responses against a 14.4% fallback rate (below the field's ~20% benchmark), 85% positive user sentiment, and a moderate positive correlation (Spearman's rho = 0.36, p = 0.012) between how specific a teacher's query was and how satisfied they were with the response; domain-specific keywords such as 'digital didactics' roughly doubled the odds of positive feedback (OR = 2.05). Teachers most often asked about digital competencies, AI and inclusion, and qualitative feedback praised the tool's timely, context-specific suggestions, while criticism centred on cases where recommendations still felt generic.

The study, published in the peer-reviewed journal Artificial Intelligence in Education (October 2025), is a single-institution, single-region pilot: it demonstrates a well-instrumented, quantitatively evaluated PD-recommendation tool, not yet a multi-institution or cross-border deployment, and it measures chatbot performance and teacher satisfaction rather than downstream classroom or student outcomes.