Creative CRAFT: A structured framework for creativity-driven prompt engineering in generative AI

Abstract

This study introduces the Creative CRAFT Framework, an advanced prompt engineering model designed to systematically enhance both the quality and creativity of outputs generated by large language models. Based on a sample of 100 participants, comparative analyses reveal that prompts created using the Creative CRAFT Framework significantly outperform traditional prompting methods across multiple dimensions of output quality. These include task relevance, structural coherence, creativity and novelty, tone fidelity, and format accuracy. The improvements in effect size range from 18.4% to 46.8%, demonstrating the framework's effectiveness in advancing prompt engineering techniques and output quality in large language models (p < 0.0001). Concurrently, user perception assessments reveal elevated levels of usability, clarity, and satisfaction, with particular emphasis on the framework’s efficacy in fostering creative expression. Thematic analysis of qualitative feedback corroborates these quantitative outcomes, elucidating the framework’s modular design, flexibility in component integration, and the critical role of the Creative Direction element in eliciting imaginative and contextually nuanced responses. The framework’s six components Context, Role, Action/Task, Format, Tone/Steps/Constraints, and Creative Direction are organized within a non-linear, circular schema that prioritizes completeness over sequential order in prompt construction. This structural configuration enables user adaptability and purposeful prompt formulation, facilitating a calibrated balance between methodological rigor and creative freedom. Collectively, these findings affirm the Creative CRAFT Framework as a significant contribution to prompt engineering, delivering a robust, user-centric methodology that enhances the expressiveness, relevance, and overall quality of AI-generated content.