GenAI is changing teachers' work faster than the field is changing the way it measures teachers' competence. Self-report scales inherited from the TPACK tradition will remain useful, and the development of comprehensive scales is part of the same agenda. If research relies on them alone, however, it is likely to describe teachers' perceptions of AI integration in increasing detail while saying comparatively little about the practice of integration. This paper is an attempt to draw attention to the need for observation-based, practice-oriented rubrics that can stand alongside these scales.
A research agenda follows from this argument. Observable indicators for AI-specific competencies need to be defined and subjected to expert review; rubrics need to be calibrated with anchor lessons and rater training; their reliability needs to be examined with rater-mediated psychometric models and generalizability studies; their relations with self-report scales, lesson plans and scenario-based tasks need to be investigated; and their functioning needs to be compared across pre-service and in-service teachers and across countries. It is recommended that journals welcome studies that combine self-report with observational, artifact-based, and scenario-based evidence; that teacher education programs assess pre-service teachers' AI integration in microteaching and practicum settings; and that researchers openly share validated rubrics for classroom use of GenAI.