ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›AI MEANINGS›WHAT IS A SYSTEM PROMPT?
AI MEANINGS · MEANINGBEGINNER

What is a system prompt? The standing instructions behind a chatbot

A system prompt is the hidden set of standing instructions an AI chatbot gets before you type. It sets its role, tone and rules. Here is how it works.

READ5 min
WORDS1,052
SECTIONS6
TYPEMEANING
CHECKED1 OCT 26

A system prompt is the set of standing instructions an AI chatbot is given before you type anything. It sets the AI’s role, tone and rules for the whole conversation.

◈ IN PLAIN TERMS

Think of a new shop assistant. Before the first customer walks in, the manager hands over a staff handbook: “You work in the garden department. Be friendly. Never promise refunds.” Customers never see the handbook, but it shapes every answer they get. The system prompt is that handbook.

In a sentence: “The bakery’s chatbot kept steering me back to cake orders, because its system prompt told it to stick to the shop’s menu.”

What the term means

It is written by whoever builds the app, and it comes first. When a company puts an AI model inside a product, it adds a block of instructions ahead of your messages. NIST, the US government’s standards agency, describes it like this: the developer or user provides instructions “often prepended to the model’s other input”, a description of the use case such as “You are a helpful financial assistant who responds gracefully and concisely....”, which “is known as a SYSTEM PROMPT.”1

“Prepended” just means placed in front. The model reads the system prompt, then your message, and answers with both in mind.

WHAT THE MODEL READS · IN ORDER
You see your message and the reply. The system prompt sits in front of them, unseen.
The system promptThe app builder’s standing instructions: role, tone, rules.
Your messageWhat you type in the chat.
The answerWritten with both in mind.
Reasoning, 1 Oct 2026 — summarises this page’s section on what the term means.

A simple example

The same model can act like two different products. One company’s system prompt says: “You are a patient tutor for ten-year-olds. Use short words. Never give the answer straight away.” Another says: “You are a coding assistant for professional developers. Be brief.” Ask both the same question and you get very different replies, even if the AI underneath is identical.

Anthropic, which makes Claude, says even one sentence of role in the system prompt changes behaviour.2

Why it matters to you

Every chatbot you use has rules you cannot see. When an app’s AI refuses a topic, sticks to one brand’s products, or always answers in a certain style, the system prompt is often why. It is not the AI’s own opinion; it is the builder’s instructions.

It is also what attackers try to override. A jailbreak or a prompt injection is, at heart, an attempt to make the model ignore its system prompt. NIST, the US standards agency, describes a prompt injection as an attack that overrides the system prompt’s instructions.3 So a system prompt guides the AI, but it is not a lock.

If you build your own assistant or agent, the system prompt is where your lasting instructions go. The agent brief gives a structure for writing one.

Often confused with

TermHow it differs
Your promptDiffers: what you type in the chat box. It changes with every message; the system prompt stays in place behind it.
TrainingDiffers: training changes the model itself. A system prompt is just text read at the start of each chat, so it can be edited any time.
MemoryDiffers: memory is a product feature that saves details about you between chats. See AI terms people get wrong.

Questions people ask

Can you see a chatbot’s system prompt?

Usually it is hidden from you, but some companies publish theirs. Anthropic posts the system prompts for its Claude apps, which give Claude information such as the current date at the start of each chat; it says these do not apply when developers use Claude through its API.4

Can a system prompt stop jailbreaks?

Not reliably. A system prompt guides the AI but is not a lock: NIST, the US standards agency, describes prompt injection as an attack that “overrides these instructions”. Prompt injection meaning explains how.3

Learn more on this site

Related on this site: How Does ChatGPT Work? · What Is Constitutional AI? · Prompt vs System Prompt · What Is Role Prompting?

QUICK CHECK · 3 QUESTIONS

Pick an answer to see if you got it. Nothing is recorded.

1. What is a system prompt?

  • The first message you type into a chat
  • Changes made to the model during training
  • Standing instructions the AI gets before you type
Show the answer

Answer: Standing instructions the AI gets before you type. The page says the system prompt is the app builder’s set of instructions, placed in front of your messages, that sets role, tone and rules.

2. Why can the same AI model act like two different products?

  • Each company trains a new model from scratch
  • Each company gives it a different system prompt
  • The AI picks a personality at random
Show the answer

Answer: Each company gives it a different system prompt. The page’s tutor and coding assistant example shows one model giving very different replies because of different system prompts.

3. Can a system prompt reliably stop jailbreaks?

  • No, it guides the AI but it is not a lock
  • Yes, the AI can never ignore it
  • Yes, as long as it is kept secret
Show the answer

Answer: No, it guides the AI but it is not a lock. The page says jailbreaks and prompt injection are attempts to make the model ignore its system prompt, so it is not a lock.

Related terms: AI literacy · Role prompting · Sentiment analysis · Sycophancy · Temperature · Zero-shot prompting

SOURCES
  1. 1NIST, AI 100-2e2025, Adversarial Machine Learning: A Taxonomy and Terminology of Attacks and Mitigations, March 2025, read at source 1 Oct 2026the developer or user provides instructions “often prepended to the model’s other input”, a description of the use case such as “You are a helpful financial assistant who responds gracefully and concisely....”, which “is known as a SYSTEM PROMPT.” ↑
  2. 2Anthropic, Prompting best practices, read at source 1 Oct 2026“Setting a role in the system prompt focuses Claude’s behavior and tone for your use case. Even a single sentence makes a difference”. ↑
  3. 3NIST, AI 100-2e2025, March 2025, read at source 1 Oct 2026“A PROMPT INJECTION overrides these instructions, exploiting the concatenation of untrusted user output to the system prompt to induce unintended behavior.” ↑a“A PROMPT INJECTION overrides these instructions, exploiting the concatenation of untrusted user output to the system prompt to induce unintended behavior.” ↑b
  4. 4Anthropic, System prompts (release notes), undated documentation, read at source 12 Oct 2026“Claude’s web interface (claude.ai) and mobile apps use a system prompt to provide up-to-date information, such as the current date, to Claude at the start of every conversation.” and “These system prompt updates do not apply to the Claude API.” ↑

The comparisons on this page (the shop assistant’s handbook, the tutor and the coding assistant) are this site’s own ways of explaining, not claims from these sources.

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE