Assistant Message

Artifact prompting published

Also known as: Model Response, AI Response, Completion

Definition

The output generated by the LLM in response to the prompt, representing the "assistant" turn in a conversation. Assistant messages contain the model's answers, generations, or actions (like tool calls). In multi-turn conversations, past assistant messages become part of the context for subsequent turns.

What this is NOT

  • Not the user message (that's input)
  • Not the system prompt (that's instructions)
  • Not training data (assistant messages are inference outputs)

Alternative Interpretations

Different communities use this term differently:

llm-practitioners

A message with role "assistant" in the Chat Completions API. Represents what the model has said or will say. Can contain text, tool calls, or both.

Sources: OpenAI Chat API documentation, Anthropic Messages API documentation

Examples

  • Assistant: The capital of France is Paris.
  • Assistant: [tool_call: get_weather(location='Paris')]
  • Assistant: Let me search for that. [tool_call: search(...)] Based on the results...
  • Conversation history showing alternating user/assistant turns

Counterexamples

Things that might seem like Assistant Message but are not:

  • User's question or instruction
  • System prompt
  • Tool result messages

Relations

  • inTensionWith user-message (Assistant vs. user is the fundamental conversation structure)
  • overlapsWith function-calling (Tool calls appear in assistant messages)
  • produces tool-result (Tool calls in assistant messages lead to tool results)

See Also