Data Science

What is Prompt Engineering? 3 Steps For Good Prompt

Automatically translated·See original

Masqot CommunityCreator

March 4, 2024

4 min read

Save

What is prompt engineering, how to write good prompt? Connect, ton and fine adjustment steps; sample prompt, hallucination alert and frequently asked questions.

Prompt engineering is to design the instruction consciously to get the result you want from an artificial intelligence chat boat. A good prompt is installed in three steps: first draw context and limits, then determine the tone and target audience, in the last example, you will finely adjust with the output format and verification request. ChatGPT, Gemini (formerly Bard) and Claude all work the same logic.

Last update: 23.10.2026

When writing an effective prompt, you need to consider many things at the same time: topic, purpose, reading, format, length. It can mix the head to set all at one time. That's why it makes it easy to split the work into three steps.

Home	Tags u	What do you add to Prompt?
1. Limits	What I want, in what context?	Task, background information, scope, output format Home
2. 2. Tone	Who, what purpose is to hear?	Target audience, purpose, tone, issues to be avoided
3. Fine adjustment	How is it better and reliable?	Samples, benzing, step-by-step thinking, verification

Prompt Engineering: Critical Step 3

1. Draw Limits (We Are We Voiceed?)

Chat bot works to predict what he wants because he is trained with a very wide knowledge. It estimates less if you give the net context. Write the task clearly, add the required background information and limit the scope. Give one or two examples that are similar to the conclusion you want if possible; the guide of OpenAI also suggests to give input-line samples (few-shot) and separation to the instruction sections. Tell the format of output: list of items, table, email, how many words?

Example: "The main substances of a remote working policy for a software company's HR team. Company 40 persons. No matter 6, each item is up to two sentences. "

2. 2. Ton and Target Audile (How Do We Voice?)

Put the rules of the game in the first step, tell how to play now. Will the tone of the output be official, intimate? Tonu determines the purpose, the purpose of the target audience. An announcement that goes to community members with a summary of the presentation to the investor is not written in the same language. It clearly states who write in Prompt (“non-technical executives”, “the students who apply for the first time” and its purpose. If you add specific terms and conditions to the subject or expressions that should be avoided, the result will be much closer to your base. It is also useful to give a role: When the "response like a trained HR consultant" the model adjusts both tone and the level of detail accordingly.

3. Fine Setting: Sample, Format and Verification

You get enough results with the first two steps; you can get much better with small additions:

1. Benzolysis: In order to understand a complex concept, "to tell you about it with a humor."

2. Step by step: In multi-step questions, ask for model to show intermediate steps (chain-of-thought) can improve results.

3. Different perspectives are asked: "In favor of this decision and against three reasons like summer".

4. In the source name but check: Chat bots can produce non-welding, quote or statistics (halüsination). Open and correct each figure and connection that it gives.

5. Personal data bonding: The customer has anonymized its member or employee information before putting it into prompt; processing personal data in Turkey is subject to KVKK, which is no. 6698.

6. Again: The first answer is to say what is missing if it did not sit instead; good prompt often goes on several rounds.

These steps are called "prompt engineering" to implement in the system. In 2023, it was very spoken as a separate profession title; but the models seem better to understand the user's intention, and as companies give them this skill to their employees, the Wall Street Journal was transferred in the news of 2025, as requested to "prompt engineer" positions. Today prompt writing is seen as a basic skill that everyone uses from software to marketer. If you are curious about other developments on the technology side Apple Vision Pro you can read. For more Masqot wait.

Frequently Asked Questions

What is prompt engineering?

The instruction (prompt) context to get the result you want from an artificial intelligence model is to design consciously with the purpose, example and output format.

How to write a good prompt?

The task is open summer, give the required context, specify the target audience and tone, tell the format you want (list, table, length) and add an example if possible. Still saying what’s missing if you don’t like the son.

Is the prompt engineering still a profession?

Separated positions were reduced by 2023; the skill has become part of many works. Prompt design and assessment in teams that improve artificial intelligence product is still an important job.

Can I trust Chatbot’s resources?

Trust without verification. Models can produce non-welding and statistics; control each link and figure from itself.

Is it a problem to write personal data to Prompt?

Can. The personal data of others anonymized before bonding; the processing of personal data is within the scope of the KVKK and you should check how the vehicle uses data.

Prompt engineering is to design the instruction consciously to get the result you want from an artificial intelligence chat boat. A good prompt is installed in three steps: first draw context and limits, then determine the tone and target audience, in the last example, you will finely adjust with the output format and verification request. ChatGPT, Gemini (formerly Bard) and Claude all work the same logic. Last update: 23.10.2026 When writing an effective prompt, you need to consider many things at the same time: topic, purpose, reading, format, length. It can mix the head to set all at one time. That's why it makes it easy to split the work into three steps. Home Tags u What do you add to Prompt? 1. Limits What I want, in what context? Task, background information, scope, output format Home 2. 2. Tone Who, what purpose is to hear? Target audience, purpose, tone, issues to be avoided 3. Fine adjustment How is it better and reliable? Samples, benzing, step-by-step thinking, verification Prompt Engineering: Critical Step 3 1. Draw Limits (We Are We Voiceed?) Chat bot works to predict what he wants because he is trained with a very wide knowledge. It estimates less if you give the net context. Write the task clearly, add the required background information and limit the scope. Give one or two examples that are similar to the conclusion you want if possible; the guide of OpenAI also suggests to give input-line samples (few-shot) and separation to the instruction sections. Tell the format of output: list of items, table, email, how many words? Example: "The main substances of a remote working policy for a software company's HR team. Company 40 persons. No matter 6, each item is up to two sentences. " 2. 2. Ton and Target Audile (How Do We Voice?) Put the rules of the game in the first step, tell how to play now. Will the tone of the output be official, intimate? Tonu determines the purpose, the purpose of the target audience. An announcement that goes to community members with a summary of the presentation to the investor is not written in the same language. It clearly states who write in Prompt (“non-technical executives”, “the students who apply for the first time” and its purpose. If you add specific terms and conditions to the subject or expressions that should be avoided, the result will be much closer to your base. It is also useful to give a role: When the "response like a trained HR consultant" the model adjusts both tone and the level of detail accordingly. 3. Fine Setting: Sample, Format and Verification You get enough results with the first two steps; you can get much better with small additions: Benzolysis: In order to understand a complex concept, "to tell you about it with a humor." Step by step: In multi-step questions, ask for model to show intermediate steps (chain-of-thought) can improve results. Different perspectives are asked: "In favor of this decision and against three reasons like summer". In the source name but check: Chat bots can produce non-welding, quote or statistics (halüsination). Open and correct each figure and connection that it gives. Personal data bonding: The customer has anonymized its member or employee information before putting it into prompt; processing personal data in Turkey is subject to KVKK, which is no. 6698. Again: The first answer is to say what is missing if it did not sit instead; good prompt often goes on several rounds. These steps are called "prompt engineering" to implement in the system. In 2023, it was very spoken as a separate profession title; but the models seem better to understand the user's intention, and as companies give them this skill to their employees, the Wall Street Journal was transferred in the news of 2025, as requested to "prompt engineer" positions. Today prompt writing is seen as a basic skill that everyone uses from software to marketer. If you are curious about other developments on the technology side Apple Vision Pro you can read. For more Masqot wait. Frequently Asked Questions What is prompt engineering? The instruction (prompt) context to get the result you want from an artificial intelligence model is to design consciously with the purpose, example and output format. How to write a good prompt? The task is open summer, give the required context, specify the target audience and tone, tell the format you want (list, table, length) and add an example if possible. Still saying what’s missing if you don’t like the son. Is the prompt engineering still a profession? Separated positions were reduced by 2023; the skill has become part of many works. Prompt design and assessment in teams that improve artificial intelligence product is still an important job. Can I trust Chatbot’s resources? Trust without verification. Models can produce non-welding and statistics; control each link and figure from itself. Is it a problem to write personal data to Prompt? Can. The personal data of others anonymized before bonding; the processing of perso

Masqot CommunityCreator

@masqot