Generative AI models can be powerful tools for legal research, drafting, and analysis. However, the quality of AI output depends on the quality of the prompts provided. A well-crafted prompt provides context (case type, facts, documents), clarity (what exactly you want the AI to do), and structure (step-by-step guidance). Think of prompting like instructing a junior lawyer: the clearer and more complete your instructions, the stronger the resulting work product.In this guide, we will begin with general prompting principles, introducing the COSTAR framework (Context, Objective, Style, Tone, Audience, Response) as a foundation. Then we move into techniques tailored for advanced models like GPT-5.1, explaining how to structure, refine, and control prompts to achieve reliable results.  After these fundamentals, the guide provides three examples of prompts in the COSTAR framework.

‍