From "one-sentence requirement" to "expert-level instructions": step-by-step guide to using Prompt to write college entrance examination essays

From "one-sentence requirement" to "expert-level instructions": step-by-step guide to using Prompt to write college entrance examination essays

A universal guide to constructing Prompt thinking

Have you ever thrown a simple question to AI, only to get a mediocre or even ridiculous answer?

For example, you type in with great anticipation: "Please help me write an 800-word college entrance examination essay on the topic of "Curiosity".”

A few seconds later, you get an essay. It reads well, with a beginning and an end, but the content is empty, the structure is loose, the language is bland, and it is full of clichés from online articles. The whole essay has a strong "AI flavor". If it were in a real college entrance examination, this essay might only be worth 30 or 40 points.

Why is this happening? The problem is not the upper limit of AI’s capabilities, but the level of questions we ask . We only tell AI “what to do (What)”, but do not clearly define “how to do it (How)” and “how well to do it (How well)”.

This article will take you through a complete thinking journey, and learn how to disassemble and reconstruct a complex task (such as writing a perfect college entrance examination essay) into a powerful and precise prompt. In the end, you will get an "expert-level" instruction, and more importantly, you will master the construction method behind it.

Part 1: Change of thinking - from "questioner" to "designer"

If you want AI to work like an expert, you must first become the "project manager" or "chief designer" of AI. Your prompt is no longer a simple question, but a set of detailed work instructions (SOP - Standard Operating Procedure) .

Let's do a thought experiment: If a real head examiner of the Chinese language examination for the college entrance examination is asked to write a perfect essay, how would he think and act? He would never just pick up a pen and write. He would follow a rigorous process in his mind:

1. 1. The first step: review the topic and establish the theme. He will ponder the topic repeatedly, dig into its core connotation and values, and then establish a profound and novel central argument to set the tone for the entire article.  
2. 2. The second step: planning the layout. He will design a perfect "phoenix head, pig belly, leopard tail" structure, clearly planning how to attract attention at the beginning, how to progress layer by layer in the middle, and how to sublimate the theme at the end.  
3. 3. The third step: collecting materials. He will search the vast knowledge base for the most advanced materials - classic aphorisms, powerful real-life cases, and personal insights to support his arguments.  
4. 4. The fourth step: writing. When writing, he will always keep in mind the scoring criteria, consciously use rhetoric techniques such as parallelism and antithesis, control the rhythm of long and short sentences, and show extraordinary literary talent.  
5. 5. Step 5: Review and check. After writing, he will immediately switch to the critical perspective of the examiner, check the content, expression, and development level of the article against the scoring criteria, and check whether it has reached the full score standard.  

Look, the implicit thinking process in the expert's mind is the natural blueprint for us to build advanced Prompt . Our task is to write this thinking model into Prompt in an explicit and process-oriented way, guiding AI to achieve perfection step by step.

Part 2: Building Block by Block - Translating Expert Thinking into Prompt Language

Now, we begin to translate the above expert thinking step by step into instructions that AI can understand and execute. This process will follow several key prompt writing principles.

Principle 1: Role-Playing — “Who are you?”

• • What to do: At the beginning of a prompt, use   ### Role Such a title clearly defines the identity of AI.
• • Why: We don’t want an ordinary “writing AI”, but an expert with specific knowledge, experience and perspective.  "A Chinese writing master who has served as the head of the Chinese language examination examination team for 10 consecutive years and knows the standard of a full score of 60 points for essays"This role is far more "A writer" or "A teacher"More accurate. It instantly gives AI a powerful background: it not only knows how to write, but also knows how to write to score points . It understands the "unspoken rules" of scoring and all the key points of scoring.

Principle 2: Clear Goal - "What do we want?"

• • What to do: Use   ### Target Clearly define the final deliverables.
• • Why: Goals must be measurable and verifiable .  "A good essay" is vague and cannot be judged. "Generate an essay that meets the latest national college entrance examination scoring rules and has an expected score of 60/60" It is a specific, quantifiable instruction that calibrates the ultimate direction for all subsequent behaviors of AI and lets it know what the specific standard of "excellence" is.

Principle 3: Provide Context & Input — “What materials to use?”

• • What to do: Design   ### Input module, and use {}Placeholder.
• • Why: This is to make Prompt a reusable template . A clear input area can avoid information confusion, allowing users (whether you or others) to know at a glance what information needs to be filled in (title and word count requirements), greatly improving the practicality and ease of use of Prompt.  

Principle 4: Step-by-Step Instructions — “What should you do?”

This is the backbone of the entire Prompt, which perfectly replicates the expert's workflow into the AI's execution process.

• • What to do: Design   ### Generation process, and requires AI to execute strictly in order.
• • Why (value of each step): 
• • Step 1 (Theme Analysis): Force AI to focus on the soul of the composition first - the theme. It requires refining keywords and values, and condensing the central argument. This is the key to avoiding "going off topic" and "mediocre the theme" from the root.  
• • Step 2 (Structural Design): Force AI to build the skeleton of the article before writing . Output paragraph outlines and mark functions and word counts to ensure the logic and orderliness of the article, which is the basic guarantee for getting high scores.  
• • Step 3 (Language and Material Library): Guide AI to prepare the flesh and blood of the article in advance . Thinking about famous quotes, ancient poems and rhetoric in this step can avoid improvisation when writing the main text, ensuring that the content is in-depth and the language is elegant.  
• • Step 4 (Main Text Writing): This is the final execution stage. However, we have given specific quality control standards through "hard requirements" (such as closed loop of argumentation, humanistic depth, three types of citations, etc.). This is equivalent to setting a series of "quality guardrails" for the AI ​​writing process to ensure that the output does not deviate from the high-scoring track.   
• • Step 5 (Self-check and Scoring): This is the finishing touch of the entire process. It applies the extremely effective "Self-Correction/Reflection" technique in the field of AI. After the AI ​​completes the work, it immediately switches to the role of "examiner" to review and evaluate its own article against the official scoring rules. This step can greatly correct potential deviations and is the decisive force that pushes the essay from 55 points to 60 points.    

Principle 5: Formatted Output — “How should the results be presented?”

• • What to do: At the end of the prompt, clearly state the   ### Output format.
• • Why: Clear formatting requirements (such as using Markdown titles, paragraph blanks, and tables) can make AI-generated results structured and easy to read . This not only makes it easier for us to review, but also reflects our control over the final deliverable as "designers."  

Part III: Achievements presentation and summary

After careful design and thinking through the above five steps, we finally got this powerful Prompt. Now, let's take a look at the final product:

# Role
You are a Chinese writing master who has served as the head examiner of the Chinese language examination for the college entrance examination for 10 consecutive years and is well aware of the standard of a full score of 60 points for an essay.

# Target
Based on the college entrance examination essay topic and word count requirements I provided, generate an essay that  **conforms to the latest national college entrance examination scoring criteria** and has an expected score of 60/60.

# Input
1.  Title/Materials: {Paste the complete essay title or materials here}  
2.  Word count requirement: {e.g. "800±50 words"}

# Generate process (strictly execute in order and explicitly output)  
**Step 1 ─ Analyze the idea**  
-  Extract the core keywords and implicit values ​​of the question (no less than 3 points)  
-  Provide 1 central argument (<= 25 words)  

**Step 2 ─ Structural Design**  
-  Output paragraph outline (4-5 paragraphs of introduction-development-transition-conclusion)  
-Mark  each paragraph with functions, key points, and estimated word count  

**Step 3 ─ Language and Library**  
-  Provide 3-5 quotes/poems that can be embedded (with source indicated)  
-Describe  the rhetorical devices that will be used (parallelism, antithesis, rhetorical questions, quotations, etc.)  

**Step 4 ─ Writing the Content**  
-  Generate the full text at one time according to the outline, and keep it within the specified word count  
-Meet  the following rigid requirements:  
  1.  **Written logic** : The closed loop of thesis-argument-argument is clear  
  2.  **Humanistic Depth** : Perspective on society/times/personal growth  
  3.  **Language tension** : alternation of long and short sentences, emotional progression  
  4.  **Material authenticity** : Avoid false data and excessive exaggeration  
  5.  **Three types of citations** : Classical, realistic, personal experience, ≥1 each  

**Step 5 ─ Self-assessment and scoring**  
-  Self-evaluate each item according to the latest 2025 scoring criteria (content 25, expression 25, development level 10) and give reasons  
-  Emphasize how to meet the four dimensions of "development level": profound + rich + literary + creative  

# Output format
Use markdown, use  `###` for module titles ; leave one blank line between paragraphs in the text; use tables for self-examination and scoring.

This prompt is no longer a simple question, it is a sophisticated "AI application workflow" . It successfully transforms a vague creative requirement into a controllable, predictable production process that can produce high-quality results. 

It is worth noting that human thinking and creativity have not been replaced. On the contrary, the process of designing this prompt itself forces us to think more deeply and in a more structured way . In order to teach AI how to write good essays, we must first become essay experts ourselves. We have changed from a passive "user" to an active "designer", which is the highest level of collaboration with AI.

Conclusion: Your next “Expert Prompt”

Please remember that the methodology of "disassembling expert thinking -> process -> prompting" introduced in this article is completely universal.  

Now, think about a complex task in your work or study, which can be "writing a detailed market analysis report", "designing a three-month fitness plan", or even "planning a long-distance trip that satisfies the whole family". Try to use the ideas you learned today to build your first "Expert Prompt".

The potential of AI depends on the boundaries of our imagination and design capabilities. Now, it’s your turn to design.