Home › Encyclopedia › Prompting Techniques

System Prompt

Latest update: 26/04/29

Definition

A system prompt is a set of instructions given to an AI before a conversation starts – usually by a developer or platform – that shapes how the model behaves, what role it plays, and what it will or won’t do throughout the session.

What Is a System Prompt?

When you open a chat interface and start talking to an AI, what you type isn’t the first thing the model reads. Before any conversation begins, most AI deployments include a system prompt – a block of instructions that sets the rules of engagement.

System prompts are written by whoever built or configured the AI tool you’re using, not by you. They might define the AI’s persona, restrict it to certain topics, give it background knowledge about a product or company, tell it what language to respond in, or establish a tone it should maintain across every interaction.

As an end user, you usually can’t see the system prompt. But you’re always operating inside its constraints. It’s the invisible stage direction that runs before the curtain goes up.

💡 How Does It Work?

System prompts sit at the top of the context window, before the conversation history. Every time the model generates a response, it reads the system prompt first – which means those instructions apply consistently throughout the session, not just for the first message.

Think of it like a job briefing. Before an employee starts their shift, they’re told: who they’re serving, what they can offer, what to do in difficult situations, and how to represent the company. The system prompt does the same thing for an AI – establishes context, role, and ground rules before any customer interaction begins.

A developer building a customer service chatbot might write a system prompt like: “You are a support agent for Acme Software. Answer only questions related to our products. Always be polite. If a question falls outside our product scope, direct the user to our help center.”

Every user conversation that follows runs within those parameters.

Why It Matters for Your Prompts

If you’re a regular user of AI chat tools, system prompts explain a lot of the behavior you might notice but can’t account for: why the AI always introduces itself a certain way, why it refuses certain topics, why it maintains a specific tone even when you try to redirect it, or why it seems to know about a specific product without you explaining it.

For developers and power users building AI applications, system prompts are one of the most important tools available. They’re where you define the AI’s behavior at a structural level – not through conversation, but through persistent instruction.

For everyday users who want more control: if you’re working directly with an API or a tool that lets you set a system prompt, this is where to put instructions that should apply across the whole session – role definitions, tone requirements, output format standards, topics to avoid. It’s more reliable than repeating those instructions in every user message.

🌐 Real-World Example

Real-World Example

A law firm builds an AI research assistant for its associates. Without a system prompt, the AI behaves as a general-purpose assistant – it’ll chat casually, offer opinions, and answer questions about anything.

With a system prompt, they configure it to: always identify itself as a legal research tool, respond only to questions about case law and legal procedure, never offer legal advice or opinions, always cite sources, and flag when a question falls outside its scope.

Associates get a focused, professionally appropriate tool. The system prompt does the configuration work once – and every conversation that follows inherits those rules automatically.

Related Terms

Prompt – The user’s message is called the user prompt; the system prompt is what comes before it and shapes how the model interprets everything that follows.
Role-Based Prompting – Assigning a persona or role to the AI is one of the most common things developers put in a system prompt.
Prompt Template – A system prompt is essentially a persistent template applied at the platform level, not the conversation level.
Context Window – System prompts consume context window space; long, detailed system prompts leave less room for conversation.
Negative Prompting – Restrictions on what the AI should never do belong in the system prompt rather than repeated in each user message.

Encyclopedia Fundamental Architecture & Technical Advanced Concepts

Frequently Asked Questions

Can I see what system prompt an AI tool is using?

Usually not. Most commercial AI products keep their system prompts confidential. Some platforms allow users to view or set their own system prompts – particularly API access and developer-focused tools. If you’re using a consumer product like Claude.ai or ChatGPT, there is a system prompt running in the background, but it’s not shown to users.

Can I override a system prompt in my messages?

Sometimes, partially – but it depends on how the system prompt was written and which model you’re using. A well-designed system prompt resists casual override attempts. A poorly written one may be redirected by a persistent or clever user message. This is also the domain of prompt injection attacks, where malicious input tries to override system instructions.

If I’m using the API directly, should I use a system prompt?

Yes, for almost any serious use case. The system prompt is the right place for instructions that should apply to every response – persona, format rules, topic scope, tone, constraints. Putting those instructions in every user message works but is wasteful and inconsistent. The system prompt handles it once.

Does the model follow system prompt instructions more strictly than user instructions?

Generally, yes – especially on well-aligned models. System prompts are treated as higher-priority configuration by design. That said, no instruction layer is absolute, and very capable models can be influenced by user-turn conversation when system prompts aren’t carefully written. The hierarchy exists, but it’s not a hard lock.

References

Anthropic – System Prompts – Official documentation on how system prompts work in Claude and best practices for writing them.
OpenAI Developer Cummunity – The system role – How it influences the chat behavior