Prompt
Definition
A prompt is the input supplied to an AI model or AI system to guide its behaviour or output. A prompt can contain a question, instruction, context, examples, constraints or structured data.
Why it matters
Prompts are the starting point for many AI interactions, but they are not necessarily the same as the retrieval queries an AI system later issues. A single prompt may be interpreted, rewritten, decomposed, routed or expanded before external information is retrieved.
How it works
User input
- A prompt can contain natural-language instructions, questions or task requirements.
- System and developer instructions may also shape model behaviour in addition to the user’s prompt.
Interpretation
- The AI system interprets the prompt to determine intent, constraints, entities and required actions.
- Some systems then generate one or more internal or provider-exposed retrieval queries.
Context
- Prompt content is processed together with other context supplied to the model.
- Retrieved evidence and tool outputs may be added after the original prompt.
How Netsleek uses the term
Netsleek distinguishes the original user prompt from query reformulation, query expansion and query fan-out. This is important when measuring AI visibility because the words a user types may not be the same words or retrieval pathways used by the system.
Comparisons
- Prompt vs Query: A prompt is the input to the AI system. A query is an information request that may be sent to a retriever or search system.
- Prompt vs Query Fan-Out: Fan-out can generate multiple retrieval queries from one originating prompt.
- Prompt vs Context: The prompt is part of the context, but context can also include prior messages, retrieved evidence and tool results.
How Prompt Connects Across AI Systems
Prompt does not operate in isolation. It connects to other established concepts across the AI search and retrieval process:
- Interpreted by query understanding: Query Understanding can extract intent, entities and constraints from the prompt.
- Can trigger reformulation: Query Reformulation and Query Expansion may transform the information need for retrieval.
- Can trigger fan-out: Query Fan-Out and Multi-Query Decomposition can generate multiple retrieval paths.
- Conditions generation: The prompt also guides the final model response.
Related glossary concepts
- Query Understanding
- Query Reformulation (Query Rewriting)
- Query Expansion
- Query Fan-Out
- Multi-Query Decomposition
- Context Window
- LLM Synthesis
Common misinterpretations
- A prompt is not automatically a search query.
- The exact user wording does not prove the same wording was used for retrieval.
- A prompt can influence generation even when no external retrieval occurs.
Summary
A prompt is the input that guides an AI model or system. In AI search, it can be interpreted and transformed into one or more retrieval queries before the final answer is generated.