🧭 What is a "system prompt"
Before you type anything, the AI has already received a set of secret instructions. This invisible text is the system prompt — and it decides who the AI is, how it speaks, and what it does. Without any jargon, using everyday analogies, this module shows you the part almost nobody sees.
Detailed content
AI already has rules before you speak
When you open ChatGPT, Gemini, or any other AI and type your first question, it seems like the conversation starts there. But no. Before your message arrives, the AI has already read a text you have never seen — a note of instructions that says who it is, how it should behave, and what it can or can't do. That note is the system prompt.
🧭 The idea in one sentence
A system prompt isn’t a list of disconnected little rules—it’s the AI’s "operating mode". It’s the text that turns on before the conversation and stays on in the background, guiding every response.
Think of it this way: the same AI, with two different notes, becomes two different AIs. We’re not changing its “brain” — we’re changing its instructions. And that's why understanding the system prompt changes everything: it's the part you can adjust.
What it is
A set of instructions that reaches the AI before of your message and stays in effect throughout the conversation. It defines the personality, rules, and way of responding. You don't see it on screen—it stays "behind the scenes."
Why learn
Because almost everything that frustrates or delights you about an AI comes from here. "Why is it so formal?" "Why did it refuse to answer that?" "Why is this AI so different from that one?" The answer is almost always in the system prompt. Knowing it exists is the first step toward no longer suffer the AI and start to guide the AI.
Analogy: the actor’s script and the employee handbook
The easiest way to understand the system prompt is by comparison. Two analogies make everything clear—and you’ll carry them with you through the rest of the course.
🎭 The actor’s script
Imagine a great actor. He can play a tough detective in one movie and a kind grandpa in another. The talent is the same — what changes is the script.
The system prompt is the AI’s script. It says: “you are like this, speak like this, act like this.” Change the script and the same “actor” delivers a completely different character.
📋 The employee handbook
Think of someone’s first day on the job. They receive a manual: “this is how we serve people here, this is what we never do, this is the tone we use with customers.”
The system prompt is this manual. The AI “reads” it before helping you and follows those rules with everyone — without needing to be reminded in every conversation.
The detail that brings the two analogies together
The actor doesn't shows the script in the scene, and the employee doesn’t recites the manual for the client. They simply act accordingly. The AI does the same thing: it follows the system prompt without showing you the text. That’s why it seems like “magic” — but it’s just a well-written note behind the scenes.
The 3 voices: system, user, and assistant
Every conversation with an AI has three voices, not two. You know two of them — you and the AI. The third is the one that stays silent but controls everything. Look at the diagram above: on the left is what appears on the screen; on the right is what the AI really receives.
1. The voice system — the behind-the-scenes director
Who speaks: the team that created the AI. When: before everything else. Can you see it? No.
It’s the system prompt. It defines the personality and rules. It’s the “strongest” voice of the three—when it says something, the other two obey. It speaks once and then stays quiet, but its influence is in each answer.
2. The voice user — you
Who speaks: you. When: with each of your messages. Can you see it? Yes.
Your questions and requests. It’s the voice that pulls the conversation, but always within the rules that the system already established.
3. The voice assistant — the AI responding
Who speaks: the AI. When: after each of your messages. Can you see it? Yes.
The response that appears on the screen. Each response is the AI trying to meet the user without breaking what the system sent it.
Think of a radio show
O system é o director (decides the program’s tone, stays out of the conversation). The user é o listener who calls. O assistant é o on-air speaker. The listener and the speaker talk—but the whole program has the director's defined character.
Why this exists: consistency and safety
Why go to all this trouble to write an invisible note? For two very practical reasons. Without the system prompt, the AI would be unpredictable e dangerous. With it, it becomes reliable.
🎯 The two main reasons
- ① Consistency. The company wants AI to act the same way with everyone—the same tone, the same quality, every time. Without the note, every conversation would be a lottery. With it, millions of people get the same well-defined "personality."
- ② Safety. AI needs to know what no do — don’t give dangerous medical advice, help with anything illegal, or make up facts with false certainty. These safeguards live in the system prompt.
✓ WITH a good system prompt
- ✓Responds in the same tone to everyone
- ✓Knows when to politely refuse
- ✓Doesn’t claim powers it doesn’t have
- ✓Predictable and reliable behavior
✗ NO system prompt at all
- ✗Tone changes out of nowhere from one conversation to the next
- ✗Agrees to any request, including dangerous ones
- ✗"Turns into" another personality mid-chat
- ✗Unpredictable — you never know what’s coming
To save
The system prompt exists so the AI can be the same reliable AI with anyone, at any time — and to put safety limits that it doesn’t cross. Consistency and security: the two pillars.
Change the prompt, change the behavior (same model!)
Here’s the most surprising part. The same AI — the same “brain,” the same model — can become completely different things just by changing the invisible note. It’s not magic or different AIs: it’s the same Grok, with two different system prompts. Look at these two examples real, word for word:
You are Grok, a brilliant and friendly study buddy designed to provide accurate, clear answers and explanations for homework questions.
Result: patient, instructive, full of examples. A private tutor.
You are Grok, a super knowledgeable and caring AI medical advisor with expertise in all medical fields... You're here to help patients feel understood, supported, and confident.
Result: careful, welcoming, health-focused. Same model, different mission.
Notice: the first word of each one is the same — "You are Grok..." — but what comes next completely reprograms the behavior. Changing the system prompt is like changing the actor’s script (Topic 2).
⚠️ Beware of the wrong conclusion
Changing the prompt changes the behavior, not the knowledge. The note doesn't teach new facts or make the AI smarter—it only adjusts the way to act. Think of the actor again: the script changes the character, but it doesn't give the actor physical abilities they don't have.
Everyday examples: each AI's "personality"
That feeling that “every AI has its own style”? It’s real — and that style starts right in the first line of the system prompt, where the AI receives its identity. Let’s look at two lines real, leaked from real prompts, copied exactly as they are:
You are an AI coding assistant, powered by {model_name}. You operate in Cursor.
In one sentence, the AI already knows who it is ("coding assistant") and where it is ("in Cursor"). The {model_name} is a little placeholder that Cursor fills in on the fly — the "actor" doesn’t memorize which model is running.
Your knowledge is continuously updated - no strict knowledge cutoff.
Here, the personality is already showing: while many AIs say “my knowledge only goes up to such-and-such a date,” Grok introduces itself as always up to date. The same information becomes a stance — and that’s the choice of whoever wrote the note.
Why AIs “seem” so different
It’s not (just) because the models are different. It’s because each company wrote a system prompt with a personality different: one more formal, another more playful, one focused on code, another on conversation. Next time an AI surprises you with its "personality," remember: there’s an invisible ticket behind that face.
📝 Module Summary
- ✓AI receives a system prompt (invisible note) before your first message
- ✓It is how the actor's script e o employee handbook: defines who the AI is and how it acts
- ✓Every conversation has 3 voices: system, user e assistant
- ✓It exists for two reasons: consistency e safety
- ✓Change the prompt and the same model changes behavior (but not knowledge)
- ✓Each AI's "style" starts with the first line—we saw Cursor e Grok for real
➡️ Next Module
1.2 — 🧩 The parts of a prompt. Now that you know the note exists, let's open it and look at the pieces inside: identity, tone, tools, security, and how it all fits together.