Teachers’ perceptions of students’ generative AI use, professional judgment, and pedagogical adaptation: a mixed-methods study

Research on generative artificial intelligence (GenAI) in education has rarely examined how teachers interpret AI-mediated student work as evidence of understanding. This study proposes an assessment inference–pedagogical adaptation framework to examine associations between perceived student GenAI use, professional judgment, and pedagogical change. A sequential explanatory mixed-methods design combined survey data from 658 senior secondary and higher education teachers, analyzed using structural equation modeling and multi-group analysis, with subsequent interviews and instructional document analysis. Perceived student GenAI use was positively associated with assessment judgment disruption and instructional interaction disruption. Both were associated with pedagogical adaptation, which showed the largest coefficient in relation to instructional transformation. Direct associations between disruption and transformation were nonsignificant for assessment judgment and small for instructional interaction. Qualitative findings highlighted difficulty inferring understanding from polished products, increased verification during classroom interaction, and redesign intended to restore process visibility. Assessment-related associations were stronger in higher education and text- and report-oriented courses, whereas interaction-related associations were stronger in senior secondary education and problem-solving and practice-oriented courses. The study advances an account of teachers' professional judgment centered on inferential transparency, distinguishes disruption from adaptation and transformation, and identifies contextual differences in these associations without establishing causal or temporal relationships.