Prompt Engineering
Definition
Prompt engineering refers to designing inputs (prompts) with which a language model solves a task as effectively as possible.
According to Liu et al., the prompt defines the task; choosing an appropriate prompt strongly affects accuracy. Brown et al. demonstrate with GPT-3 that tasks and examples can be specified purely via text without gradient updates or fine-tuning.
Simply explained
Like a precise cooking recipe, you describe the task, ingredients, and desired outcome; then you check whether the output fits.
Why it matters in practice
Clear success criteria and recurring test tasks make prompt changes verifiable. Formulations can thus be improved based on results rather than just intuition.
Example from business
Possible example: salespeople use ready-made prompt templates to quickly create personalized offer letters.
Limits and pitfalls
- Without success criteria and tests, improvement remains unclear. Prompting does not solve every cost or runtime problem; even experienced professionals do not always find the best formulation manually.
Origin & history
The GPT-3 paper by Brown et al. appeared in 2020. Liu et al. classified prompt engineering in 2021 within research on prompt-based learning. A clear year of origin cannot be derived from this.
Use in mid-sized companies
- Possible applications: standardized templates for HR, support, and marketing, whose results are checked before use.
Frequently asked questions
Is prompt engineering the same as fine-tuning?
No; Brown et al. point out that task-specific fine-tuning requires datasets with thousands to tens of thousands of examples.
Are prompts always designed manually?
No, Liu et al. also describe methods where an algorithm searches for the best template.
Do prompts always consist of readable text?
No, Liu et al.'s survey also describes continuous prompts that are inserted as trainable variables into the embedded input.
Sources
- Language Models are Few-Shot Learners · Brown et al. (2020): Language Models are Few-Shot Learners
- Prompt engineering overview · Anthropic (o. J.): Prompt engineering overview
- Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing · Liu et al. (2021): Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing
Sources checked on 14.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: