Reviewable Japanese Can-Do teaching paths with AIŠpica, Dragana (Avtor) Perak, Benedikt (Avtor) Japanese L2Can-Do descriptorsCEFRknowledge graphlearning pathCan-Do descriptors support CEFR-aligned language education, but inventories alone do not indicate how classroom activities should be sequenced. This paper presents a procedure developed within an AI language tutor for Japanese. Teaching Can-Dos are represented by level, topic, domain, vocabulary, formulaic expressions, and grammar. Similarity measures identify related descriptors, while pedagogical constraints organize them into teacher-reviewable teaching paths. Each step generates a four-stage micro-lesson comprising content, comprehension, production, and interaction. A restaurant-ordering case, automated validation, and expert-informed reviews demonstrate the procedure’s potential for traceable, inspectable Can-Do-based lesson planning while highlighting limitations in path sequencing, content validation, and artifact management.20262026-10-29 13:34:47Članek v reviji190495ISSN pri članku: 2232-3317DOI: 10.4312/ala.16.2.197-236sl