User Message

Artifact prompting published

Also known as: User Input, User Query, Human Message

Definition

A message in the conversation from the human user (or the application on behalf of the user), as opposed to system messages or assistant responses. User messages are the primary input mechanism for LLM interactions, containing questions, instructions, or content for the model to process.

What this is NOT

  • Not the system prompt (that's developer instructions)
  • Not the assistant response (that's model output)
  • Not tool results (those have their own role)

Alternative Interpretations

Different communities use this term differently:

llm-practitioners

A message with role "user" in the Chat Completions API. User messages represent what the human (or application acting as human) is asking or telling the model.

Sources: OpenAI Chat API documentation, Anthropic Messages API documentation

Examples

  • User: What's the weather in Paris?
  • User: Summarize this document: [document text]
  • User: Code a function that reverses a string
  • Multi-turn: User asks question, assistant answers, user follows up

Counterexamples

Things that might seem like User Message but are not:

  • System prompt setting behavior
  • Assistant's response to the user
  • Tool result messages

Relations

  • specializes prompt (User messages are part of the full prompt)
  • inTensionWith system-prompt (User vs. system is a key distinction)
  • overlapsWith prompt-injection (User messages are the vector for injection attacks)

See Also