Theoretical/pedagogical background. Generative AI now permeates popular-music creation, yet the “one-click generation” paradigm risks undermining core educational objectives like creative agency, critical listening, and aesthetic judgment. In higher education, AI becomes most productive when repositioned from an endpoint to a scaffold supporting human musical reasoning.Aim/focus. This paper investigates how an AI-integrated course design can foster aesthetic and critical judgment while maintaining authorship transparency and ethical practice in AI-assisted popular-music composition.Method/approach. Across the 2024 and 2025 academic years, iterative offerings of an undergraduate popular-music composition course implemented a four-phase learning architecture: (1) Ideate & Prompt (problem framing, reference analysis, prompt design); (2) Create: Lyrics, Melody, and Structure (human-led composition with selective AI assistance); (3) Audit & Explain (interrogating model outputs for coherence, bias, provenance); and (4) Produce & Refine (arrangement, mixing, mastering with version control). Students employed text-, audio-, and MIDI-based generative AI tools throughout the creative pipeline. Evidence included (a) student portfolios (prompts, stems, logs); (b) rubric-based evaluations by instructors and peers (assessing originality, coherence, harmonic control, production quality, authorship transparency); (c) external results from an international AI songwriting competition; and (d) reflective memos. Rubric data were analyzed descriptively; reflections were thematically coded.Results/main ideas. Students progressively reframed one-click AI outputs as raw materials within a human-led workflow. Their prompt literacy improved significantly, alongside a clearer articulation of aesthetic rationales and better control over arrangement and production. The curriculum’s effectiveness received strong external validation: multiple student works created within this framework won high-level awards in an international AI songwriting competition. The competition attracts global participants from universities and professionals and is adjudicated by an international committee of renowned music-technology scholars and artists. This external success corroborated the systematic classroom evidence. Furthermore, the “authorship transparency” rubric dimension anchored practical discussions on disclosure, credit, and dataset awareness.Conclusions/implications. Grounded in successful course implementation, this study provides a comprehensive framework comprising a transferable assignment sequence, an aesthetic-judgment-centered rubric, and implementable attribution policies. Findings demonstrate that AI-assisted composition within this framework can lower technical barriers without compromising artistic standards. The design principles, tools, and insights derived offer directly referential resources for faculty development, curriculum design, and program-level assessment in higher music education.