System message

A system message is the highest-priority instruction message in a chat-style API call — used to set role, constraints, and behaviour for the entire conversation.

Most chat APIs in 2026 (OpenAI, Anthropic, Google, Mistral) accept a separate system message that the model treats as higher-trust than user turns. The system message sets persona, constraints, format rules, refusal policy, and tool guidance. It is the most cost-effective behaviour-steering surface in production LLM apps — typically more impactful than fine-tuning for the same effort. Best practices in 2026: keep it concise (under 2,000 tokens to maximise prompt caching), declarative (always / never rules), example-light (examples go in user turns), and version-controlled.

When to use system message

• Any production chat or agent feature.
• Locking output format across many user inputs.
• Enforcing safety rules.

Common mistakes

• Cramming examples into the system message — they belong in user turns.
• Variable content (timestamps, user ID) in the system prefix — breaks prompt caching.

FAQ

Related terms

• System prompt — A system prompt is the high-priority instruction block that defines a model's role, constraints, and default behaviors for an entire conversation.
• Prompt caching — Prompt caching reuses the model's internal state for a repeated prompt prefix so the API charges and computes the prefix only once across many calls.
• Instruction hierarchy — Instruction hierarchy is a model's trained ordering of trust — system prompt outranks user message which outranks retrieved content — used to resist prompt injection and jailbreak attempts.

Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/system-message.md.