ACT

ACT is a simple prompt pattern for turning ChatGPT from a text generator into a coach, trainer, or practice partner. Instead of just asking for information, you ask the AI to play a role, place you in a situation, and then make you do something with what you learned. Fancy concept, very practical result.

If regular prompting is asking for a map, ACT is asking for a driving lesson.

What the template means

Why this template works

ACT works because learning sticks better when it becomes interactive.

A lot of beginners write prompts like, “Explain budgeting,” or “Teach me job interviews.” That can work, but it often stays passive. You read, nod, maybe feel productive, and then immediately forget half of it like a password you were sure you would remember.

ACT fixes that by adding three useful ingredients:

1. Role gives the AI a job.
   Instead of generic output, you get answers from the perspective of a coach, interviewer, or trainer.

2. Scenario gives the AI context.
   Instead of broad advice, you get help inside a realistic situation.

3. Practice makes you participate.
   The AI stops being a lecturer and starts being a sparring partner.

This is especially useful for:

• interview practice
• language learning
• customer service training
• sales conversations
• difficult workplace discussions
• studying for tests
• learning software or processes

In short, ACT helps you learn by doing, not just by reading.

3 sample prompts

Sample prompt 1: Formal version

ASSUME ROLE: Act as a professional interview coach for entry-level project management jobs.
CREATE SCENARIO: I have an interview tomorrow for a project coordinator position at a mid-sized tech company. Simulate a realistic interview with 5 questions, one at a time.
TEST THE USER: After each of my answers, briefly evaluate my response, tell me one thing I did well, one thing to improve, and then ask the next question.

How the template shows up:

• A, Assume role: “Act as a professional interview coach for entry-level project management jobs.”
• C, Create scenario: “I have an interview tomorrow for a project coordinator position at a mid-sized tech company.”
• T, Test the user: “Simulate a realistic interview with 5 questions, one at a time,” plus evaluation after each answer.

Why it improves the prompt: This prompt is strong because it does not just ask for interview tips. It creates an actual practice session. The AI knows who it is, what situation to simulate, and how to interact with the user. That makes the output structured, useful, and much closer to real preparation. Also, asking for one question at a time prevents the AI from dumping a wall of text on you like a fax machine from 1994.

Sample prompt 2: Semi-structured version

Be a patient Spanish tutor.

Scenario: I am traveling in Madrid next month and need to practice ordering food in a restaurant.

Give me a short roleplay where you act as the waiter. Ask me questions in simple Spanish, wait for my answer, then correct my mistakes gently and explain the better phrasing in English.

How the template shows up:

• A, Assume role: “Be a patient Spanish tutor.”
• C, Create scenario: “I am traveling in Madrid next month and need to practice ordering food in a restaurant.”
• T, Test the user: “Ask me questions in simple Spanish, wait for my answer, then correct my mistakes gently.”

Why it improves the prompt: This version feels more natural than the fully labeled format, but it still uses the ACT structure. The role keeps the tone supportive. The scenario keeps the language focused on a real need instead of random vocabulary. The testing element turns it into active practice, which is where the actual learning happens. Reading a list of restaurant phrases is fine. Using them under pressure is better.

Sample prompt 3: Natural version

Pretend you’re a hiring manager and I’m applying for a customer support job. Put me in a realistic phone screening, ask me questions one at a time, and challenge me a little if my answers are weak. After we’re done, tell me what I handled well and what I should practice.

How the template shows up:

• A, Assume role: “Pretend you’re a hiring manager.”
• C, Create scenario: “I’m applying for a customer support job” and “Put me in a realistic phone screening.”
• T, Test the user: “Ask me questions one at a time” and “challenge me a little if my answers are weak.”

Why it improves the prompt: This prompt sounds casual, but it still has all three ACT ingredients. That is the beauty of the framework. You do not have to sound robotic to get a good result. The AI has a clear identity, a believable setting, and a practical task for the user. That usually produces better engagement, better feedback, and better skill-building than a vague request like, “Help me prepare for an interview.”

Warnings and extra tips

ACT is useful, but it is not magic. A few things to watch for:

• Do not make the role too vague.
  “Be an expert” is fuzzy. “Be a supportive algebra tutor” or “Be a skeptical hiring manager” is better.

• Do not make the scenario too thin.
  If the situation matters, include the details. What kind of job? What kind of customer? What level of learner? What setting?

• Do not forget the test.
  If you skip the practice part, ACT becomes more like roleplay theater than skill-building. Fun, maybe. Effective, less guaranteed.

• Keep expectations realistic.
  AI can simulate interviews, conversations, and quizzes well, but it is not a licensed teacher, therapist, or hiring committee. Use judgment.

• Ask for pacing.
  If you want a back-and-forth session, say “one question at a time” or “wait for my reply before continuing.” Otherwise, the AI may sprint ahead like an overcaffeinated trainer.

• Specify feedback style.
  If you want gentle correction, strict scoring, or concise coaching, say so. Different learning moods need different prompt settings.

• ACT is not enough for every task.
  If you also need formatting, length limits, audience targeting, or examples, add those too. ACT handles interaction well, but it does not replace every other prompt ingredient.

Quick summary

ACT stands for Assume role, Create scenario, Test the user.

Use it when you want ChatGPT to do more than explain. Use it when you want practice. It gives the AI a role, puts you in a realistic situation, and makes you respond, not just read.

A good shortcut to remember is this: who is the AI, what situation are we in, and how will I be tested?

Try ACT next with something real from your life: a job interview, a customer conversation, a language exercise, or a tricky workplace discussion. Pick one situation, write one ACT prompt, and let the AI coach you through it. That is where prompting starts to feel less like typing and more like training.