This paper presents a user-centered investigation into the need for structured output constraints in large language models (LLMs). The authors surveyed 51 industry professionals to identify real-world use cases and motivations for applying output constraints. They identified 134 use cases across two levels: low-level constraints ensuring structured formats and appropriate lengths, and high-level constraints requiring semantic and stylistic guidelines without hallucination. The study highlights the benefits of output constraints, including streamlining development workflows, improving user experience, and enhancing integration with existing systems. The authors also discuss user preferences for expressing constraints, finding that graphical user interfaces (GUIs) are preferred for low-level constraints, while natural language is preferred for high-level constraints. Based on these findings, the authors present CONSTRAINTMAKER, a prototype tool that enables users to experiment with and apply output constraints. The tool allows users to specify different types of output constraints and provides a visual interface for testing and applying these constraints. The study concludes that structured output constraints are essential for improving the usability and effectiveness of LLMs in real-world applications.This paper presents a user-centered investigation into the need for structured output constraints in large language models (LLMs). The authors surveyed 51 industry professionals to identify real-world use cases and motivations for applying output constraints. They identified 134 use cases across two levels: low-level constraints ensuring structured formats and appropriate lengths, and high-level constraints requiring semantic and stylistic guidelines without hallucination. The study highlights the benefits of output constraints, including streamlining development workflows, improving user experience, and enhancing integration with existing systems. The authors also discuss user preferences for expressing constraints, finding that graphical user interfaces (GUIs) are preferred for low-level constraints, while natural language is preferred for high-level constraints. Based on these findings, the authors present CONSTRAINTMAKER, a prototype tool that enables users to experiment with and apply output constraints. The tool allows users to specify different types of output constraints and provides a visual interface for testing and applying these constraints. The study concludes that structured output constraints are essential for improving the usability and effectiveness of LLMs in real-world applications.