What is Prompt engineering?

Prompt engineering is the practice of writing inputs to AI models in ways that reliably produce desired outputs. The discipline emerged from research on large language models and has since expanded to image generation (image-prompt-engineering), video generation, agentic AI, and multimodal systems. Effective prompt engineering combines three skills: clearly specifying the desired output (subject, composition, style, constraints), understanding model-specific behavior (what each model interprets reliably vs unreliably), and iterative refinement (changing one variable at a time to learn the model's response surface). Across image and video models, common prompt-engineering primitives include: positive prompts (what should be in the scene), negative prompts (what should not), composition tokens (camera angle, shot type, distance), style tokens (era, photographer, medium), and reference conditioning (uploaded images, IP-Adapter inputs, ControlNet maps). For commercial creative workflows, prompt engineering matters less than people think — the higher-leverage layer is workflow design: structured prompt libraries, brand-bible inputs, scene presets, and reference-image registries that encode brand judgment so the model can be steered with one click rather than a 200-token incantation.