WER2026 - 29th Workshop on Requirements Engineering


Use of LLMs in Requirements Engineering: An Exploratory Study in Industry

Fabio Levy Siqueira; Caio Moussa Assumpção; João Marcos de Mattos Barguil; Julio Cesar Sampaio do Prado Leite

10.29327/1884517.29-17

PDF Scholar

Abstract

Large language models (LLMs) can support requirements engineering activities in different ways. Prior studies have investigated this potential through surveys and interviews with practitioners, focusing on how LLMs are reportedly used in RE tasks. However, these studies do not examine the actual prompts employed by software engineers in real-world contexts. To address this gap, this exploratory study analyzes a log of interactions between software engineers and LLMs during the execution of requirements engineering activities. These interactions are compared with the most frequently reported uses of LLMs in requirements engineering, as identified through queries to Gemini, ChatGPT, and Claude. Our motivation is to gain insights into how practitioners are using LLMs in their requirements engineering practice. The results indicate that software engineers issue a wide variety of requests to LLMs, many of which diverge from the commonly reported usage patterns. Furthermore, the identified queries span all requirements engineering activities, suggesting a broader and more heterogeneous use of LLMs.

Keywords: LLM; requirements engineer; chat