This chapter provides a practice-oriented overview of prompt engineering for the targeted control of large language models in scientific contexts. It demonstrates how the structured design of prompts—through precise instructions, specific contexts, and clear requirements—improves the quality, relevance, and traceability of AI-generated responses. Methods are described that enable the model to understand tasks based on examples, analyze complex problems step by step, compare different solution approaches, and make its own reasoning process transparent. In addition, it explains how external information, verification steps, and the targeted parameterization of response style and length can further optimize results. The chapter also examines the role of integrated tools such as internet search, code generation, and agent-based workflows in expanding AI capabilities. A particular focus is placed on “prompting for prompts”: in this approach, the AI itself is used to develop and optimize prompts. Overall, the chapter conveys the methodological foundations for the effective use of prompt engineering in scientific work.