Adaptive Prompting and Reflective Practice in Human-AI Co-Creation

Abstract

Generative artificial intelligence (AI) is increasingly reshaping creative media production by introducing new forms of collaboration between human practitioners and AI systems. Although existing studies often describe AI as a creative partner, limited research has examined how human–AI co-creation operates within real production environments. This study adopts a practice-based research approach to investigate the development of three AI-assisted social media advertisement videos for Oppo. Data were collected from prompt sequences, production artefacts, workflow documentation, and reflective production records, and analysed using qualitative thematic analysis. The findings show that AI-assisted production is an iterative and non-linear process in which creative decisions emerge through ongoing evaluation, experimentation, and adaptation. Common challenges included unstable motion, visual inconsistencies, and unintended object behaviour, often requiring multiple rounds of prompt revision and regeneration. Three prompting strategies were identified: constraint-based prompting, corrective prompting, and refinement prompting. These strategies helped practitioners manage instability, improve output quality, and maintain creative control during production. The study further suggests that prompting extends beyond a technical instruction process and functions as a reflective, creative practice through which practitioners evaluate, adjust, and refine AI-generated outputs. Rather than replacing creative labour, generative AI operated as an assistive system that relied on sustained human judgement and intervention. The study contributes empirical insight into the hybrid and process-driven nature of human–AI co-creation in contemporary creative media production.