Title:Green Prompt Engineering: Investigating the Energy Impact of Prompt Design in Software Engineering

Abstract:Language Models are increasingly applied in software engineering, yet their inference raises growing environmental concerns. Prior work has examined hardware choices and prompt length, but little attention has been paid to linguistic complexity as a sustainability factor. This paper introduces Green Prompt Engineering, framing linguistic complexity as a design dimension that can influence energy consumption and performance. We conduct an empirical study on requirement classification using open-source Small Language Models, varying the readability of prompts. Our results reveal that readability affects environmental sustainability and performance, exposing trade-offs between them. For practitioners, simpler prompts can reduce energy costs without a significant F1-score loss; for researchers, it opens a path toward guidelines and studies on sustainable prompt design within the Green AI agenda.