AI Glossary Term

What does “Prompt” mean?

Definition

The input instructions and context provided to a generative model.

Browse the full AI GlossaryExplore all free AI guides

How Prompt works

Diagram for Prompt, layers from top to bottom: System instructions: role and rules; Context: documents, examples; Your question or task; Model's response.System instructions: role and rulesContext: documents, examplesYour question or taskModel's response

Example

A prompt can be as short as 'Summarize this article' or as long as a page with a role, background documents, examples, and a requested format. The same model can give very different results depending on how it is written.

Why it matters

The prompt is the main way you steer a generative model. Clear prompts usually produce clearer results.

Common questions

Does wording matter?

Yes. Specific, clear instructions with the needed context usually beat vague requests.

Are prompts private?

That depends on the service. Check the provider's data policy before pasting sensitive material.

Is a prompt only text?

Not always. Many models accept images, audio, or files as part of the prompt.

Related terms

In-Context LearningA model's ability to follow patterns from examples provided directly in the prompt.Prompt InjectionAn attack pattern where malicious instructions are inserted into model inputs or retrieved content.Prompt TemplateA reusable prompt pattern with variables, formatting rules, and task-specific instructions.Context WindowThe maximum amount of input tokens a language model can process at once.Prompt EngineeringDesigning prompts to improve output quality, reliability, and controllability.Parent-Document RetrievalA retrieval pattern that searches small chunks but returns their larger parent documents for richer context.

Learn more in our free guides

Prompt EngineeringAI Prompt SecurityPlan-and-Solve PromptingIP-Adapter for Image Prompts

Test yourself

Prompt Engineering QuizAI Prompt Security QuizPlan-and-Solve Prompting Quiz

See also

PretrainingPrecisionPipelinePruningPerplexityQuantizationParameter-Efficient Fine-Tuning (PEFT)RAG (Retrieval-Augmented Generation)