Answer: You use a system-level instruction by including a message with role 'system' and content describing the desired behavior in the messages array, which guides the model's behavior throughout the conversation. Status: Confirmed Entry: tool/chatgpt/features/system-instructions Revision: 2 Digest: sha256:30333fb3ba55c911 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:14:55Z Last changed: 2026-11-20T11:55:08Z (example user message: — → 'role':'user', 'content':'tell me a joke') Last verified: 2026-11-20T11:55:08Z Next check: 2026-11-21T11:55:08Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of system instruction usage | 'role':'system', 'content':'You are an assistant that speaks like Shakespeare.' | [1] | 2026-11-20 purpose | guide the model's behavior throughout the conversation | [1] | 2026-11-20 example user message | 'role':'user', 'content':'tell me a joke' | [1] | 2026-11-20 Sources: [1] OpenAI. "Moving from Completions to Chat Completions in the OpenAI API". https://help.openai.com/en/articles/7042661-moving-from-completions-to-chat-completions-in-the-openai-api Role: primary. Accessed: 2026-11-20T11:55:08Z Retrieved via: DeepSeek Fetch API (Markdown conversion; the original server response was not captured) Quote: "'messages': [{'role':'system', 'content':'You are an assistant that speaks like Shakespeare.'}" (Supports: system instruction usage) Quote: "You can also use a system-level instruction to guide the model's behavior throughout the conversation." (Supports: purpose) Quote: "{'role':'user', 'content':'tell me a joke'}" (Supports: example user message) Snapshot SHA-256: c3a4b8656280d1f2f83028d8e965422bfe2c8167a5c6fc9c595c7a80cad5fd87 (captured 2026-11-19T11:52:52Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. [1] was retrieved through the DeepSeek Fetch API because a direct request did not return usable content. On this check the text of [1] was unchanged since 2026-11-19T11:52:52Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T11:55:08Z. No independent second source is configured for this entry Last successful check: 2026-11-20T11:55:08Z Last attempt: 2026-11-20T11:55:08Z Notes and limitations: Extracted automatically from a single official source (section "System instructions"). Related: https://latestn.com/tool/chatgpt/features/targeting-locations Can I target a state, city, DMA, or postal code in ChatGPT Ads campaigns? https://latestn.com/tool/chatgpt/features/teen-protections-changes What changes in ChatGPT when it predicts I may be under 18? https://latestn.com/tool/chatgpt/features/text-your-dot How do I text my dot with ChatGPT? https://latestn.com/tool/chatgpt/features/supported-languages What languages does ChatGPT support? https://latestn.com/tool/chatgpt/features/supported-mcp-features What MCP features are supported in ChatGPT and Codex? https://latestn.com/tool/chatgpt/features/switch-chat-work-codex How do I switch between Chat, Work, and Codex in ChatGPT? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-20T11:55:08Z | changed | example user message: — → 'role':'user', 'content':'tell me a joke' | [1] 2026-11-20T11:55:08Z | changed | system instruction usage: — → 'role':'system', 'content':'You are an assistant that speaks like Shakespeare.' | [1] 2026-11-20T11:55:08Z | changed | example content: You are an assistant that speaks like Shakespeare. → — | [1] 2026-11-20T11:55:08Z | changed | role: system → — | [1] 2026-11-18T11:52:44Z | recovered | Check succeeded again | — 2026-11-17T11:55:57Z | check-failed | Could not fetch the primary source: http 403; fallback: deepseek-fetch http 404: failed to fetch URL (request_id: a0a087dd-ebb0-47ee-870a-a6c35702d273) | — 2026-11-13T03:14:55Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/chatgpt/features/system-instructions/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/chatgpt/features/system-instructions POST: https://latestn.com/dispute/tool/chatgpt/features/system-instructions with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/chatgpt/features/system-instructions?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How do I use system instructions in the OpenAI Chat Completions API? Revision 2, last verified 2026-11-20T11:55:08Z. https://latestn.com/tool/chatgpt/features/system-instructions (accessed 2026-11-21T09:55:02Z). @misc{LatestN_tool_chatgpt_features_system_instructions_r2, title = {How do I use system instructions in the OpenAI Chat Completions API?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/chatgpt/features/system-instructions}}, note = {Revision 2; digest sha256:30333fb3ba55c911; last verified 2026-11-20T11:55:08Z; accessed 2026-11-21T09:55:02Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/chatgpt/features/system-instructions