To accurately convey to AI that it is a "high-quality service," it is important to incorporate four elements into a single prompt: setting the role of the expert, presenting specific evaluation criteria, sharing ideal examples, and providing step-by-step feedback. In this article, we will explain practical techniques for helping generative AI accurately understand the value of your company's services and produce high-quality descriptions and plans, along with comparisons of bad and good examples across seven perspectives.
Three Principles for Conveying High Quality to AI
The basics of conveying high quality to AI are to clearly state the three points of "role, criteria, and examples." By providing AI with concrete judgment materials rather than abstract "high quality," the accuracy of output improves.
- Assigning a Role: Set the persona of an expert for the AI
- Clarifying Criteria: Define what constitutes high quality in numerical or conditional terms
- Sharing Examples: Show samples of the ideal tone and structure
By combining these three elements, it is believed that AI will find it easier to deeply understand the context of high-quality services.
Why Can't "Make it High Quality" Alone Convey the Message?
The reason that the instruction "Make it high quality" alone does not convey the message is that the definition of "high quality" is vague and abstract for AI.
AI excels at learning specific patterns rather than abstract terms. Even the same "high quality" can refer to logic, politeness, or expertise depending on the context. Increasing the resolution of the instructions is the first step toward improving response accuracy. If you want to check the citation situation in AI searches, you can refer to How to Check Citation Situations in AI Searches.
Tip 1: Give AI the Role of an "Expert"
By giving AI the role of an expert, you can elicit high-quality tones and ideas from that perspective. Role assignment is the most effective method among the three principles.
When a role is assigned, AI is thought to prioritize referencing the vocabulary, logic, and considerations specific to that profession from its vast training data.
Bad Example of Role Assignment
- "Explain this service in a high-quality manner."
The subject and perspective are vague, making it impossible for AI to determine the criteria for writing.
Good Example of Role Assignment
- "You are a concierge at a luxury hotel. Please explain the appeal of this service using polite language that resonates with the customer."
- "You are a top-notch copywriter. Please create an introduction that succinctly conveys the strengths of the service."
By clearly stating the role in one sentence, the tone of the output changes significantly.
Tip 2: Establish Specific Evaluation Criteria and Constraints
By presenting evaluation criteria in bullet points, AI can clearly understand "what needs to be met for high quality." The key is to replace vague adjectives with verifiable conditions.
If you break down the desired quality into 3 to 5 items and present them, AI will optimize its output to meet each item.
Four Effective Evaluation Criteria
- It is logical and written from the conclusion
- It explains technical terms in an easy-to-understand manner
- It uses polite expressions that resonate with the customer's emotions
- It is kept to about 400 to 800 characters
Examples of Specifying Constraints
- "The target audience is business people in their 40s."
- "Please avoid negative expressions and use trustworthy language."
With each added constraint, the range of AI's output is narrowed, stabilizing accuracy.
Tip 3: Share "Ideal Examples (Samples)"
When you share ideal examples, AI can accurately reproduce specific patterns rather than abstract instructions. This is known as Few-shot prompting.
By inputting past successful examples or texts with the desired taste and instructing AI to align with them, the variability of output decreases.
Examples of Few-shot Prompting
- "Analyze the tone & manner and structure of the following [successful examples] and create a description of the current service with the same quality."
- "Please write in the style closest to the following three introduction texts."
Presenting 1 to 3 examples makes it easier for AI to extract commonalities.
Tip 4: Clearly State "What You Don't Want"
By clearly stating elements to avoid, AI will steer clear of expressions that lower quality. This is known as negative prompting.
By conveying not only "what you want" but also "what you don't want," the direction of the output is doubly fixed.
Examples of Negative Prompting
- "Please avoid exaggerations and baseless assertions."
- "Avoid listing technical terms and write in plain language."
- "Please do not include comparisons with other services."
By adding about three prohibitions, the need for unnecessary revisions decreases.
Tip 5: Specify the Output Format
When you specify the output format, AI presents information in an organized manner at a high quality. Specifying the format is a practical tip that complements the three principles.
If you decide on the form of output, such as "bullet points," "table format," or "with headings," in advance, it reduces editing costs in later stages.
Examples of Format Specification
- "Use headings and bullet points, and write the conclusion first."
- "Summarize the benefits in a table format with three points."
- "Output with an introduction of 200 characters + body structure."
Tip 6: Provide Step-by-Step Feedback
By providing step-by-step feedback, you can refine the output to a quality that may not be achieved in one go. This is a mindset that emphasizes the collaborative process with AI.
Instead of expecting perfection from a single instruction, conveying specific points for correction through dialogue ultimately influences the final quality.
Good Examples of Feedback
- "This part is good, but the expression is too stiff, so please revise it to be a bit more approachable yet still elegant."
- "The basis in the second paragraph is weak, so please add a specific number."
By clearly stating "what to correct and how," AI can accurately re-output.
Tip 7: Have AI Rewrite the "Perfect Prompt"
By having AI rewrite the prompt itself, the quality of the instruction improves, leading to high-quality responses. This is a method that does not aim to complete the prompt in one go.
By passing AI the instructions you wrote off the top of your head and asking, "Please rewrite this instruction into a clearer and more structured prompt," any missing conditions will be supplemented.
Three Steps for Prompt Rewriting
- First, write the instructions as they come to mind.
- Request AI to "structure and improve this prompt."
- Use the improved version suggested by AI as the basis for the final instructions.
Comparison Table of Instruction Techniques for Conveying Services to AI
We have organized the difficulty and effectiveness of each technique. When creating descriptions of your company's services, combining multiple techniques enhances effectiveness.
| Technique | Difficulty | Main Effect | Recommended Situations |
|---|---|---|---|
| Setting Expert Roles | Low | Optimizing Tone | When you want to refine the writing style |
| Presenting Evaluation Criteria | Medium | Clarifying Quality Conditions | When you want to stabilize quality |
| Sharing Ideal Examples | Medium | Reproducing Patterns | When you want to align the taste |
| Negative Specification | Low | Avoiding Unwanted Expressions | When you want to prevent incorrect directions |
| Format Specification | Low | Organizing Information | When you want to streamline editing |
| Step-by-Step Feedback | Medium | Refining Quality | When you want to enhance completeness |
| Prompt Rewriting | Medium | Optimizing Instruction Text | When instructions are unclear |
The overall strategy for conveying your company's value to AI in the B2B domain is explained in detail in the Complete Guide to LLMO for B2B Companies.
Example of a Prompt Template for Creating Service Descriptions
Here is a basic template that can be used as is. It integrates the four elements of role, criteria, examples, and format into one structure.
You are a top-notch copywriter.
Please create a high-quality introduction for the following service.
【Service Overview】(Insert your company information here)
【Target】(Specify the customer segment)
【Conditions for High Quality】
・Write from the conclusion
・Explain technical terms in an easy-to-understand manner
・Resonate with the customer's emotions
【What to Avoid】Exaggerations and comparisons with other companies
【Output Format】Headings + bullet points, approximately 400 to 800 characters
By simply inserting your company's unique information into this template, you can more easily obtain high-quality output. For insights on improving sites that are cited by AI, refer to Site Improvement Strategies for AI Citations.
What is the Role of Humans in Final Checks of AI Output?
The quality of the text generated by AI is completed through final checks by humans. The accuracy of AI responses depends on the "resolution of instructions" and "quality of dialogue," and it is believed that the final valuation should be conducted by humans.
- Check for any factual inaccuracies
- Verify that it aligns with the actual situation of the company
- Review to ensure uniqueness is not lost
AI serves only as an auxiliary role, and human adjustment and oversight are essential for quality assurance. The overall success points for utilizing AI can be checked in Success Points for AI Business Utilization.
Frequently Asked Questions (FAQ)
Q1. Is it insufficient to just tell AI "Make it high quality"?
It is insufficient. The definition of "high quality" is vague for AI, so it is necessary to present specific evaluation criteria such as logic, politeness, and expertise in 3 to 5 items.
Q2. How should I write role assignments specifically?
Specify the profession or position in one sentence, such as "You are a top-notch [profession]." The clearer the role, like "concierge at a luxury hotel," the more stable the tone will be.
Q3. How many ideal examples should I present?
1 to 3 examples are a good guideline. Presenting multiple examples makes it easier for AI to extract common patterns, reducing variability in output.
Q4. What is negative prompting?
It is a technique that clearly states "what you don't want." Adding about three prohibitions, such as "avoid exaggerations" or "do not include comparisons with other companies," reduces unnecessary revisions.
Q5. How much should I specify the output format?
It is efficient to specify a format that makes later editing easier, such as "with headings," "in bullet points," "in table format," or "approximately 400 to 800 characters."
Q6. What should I do if I don't get a good response in one go?
Provide step-by-step feedback. Clearly communicate "what to correct and how," and engage in dialogue to refine the quality.
Q7. What does it mean to have AI rewrite the prompt?
This method involves passing AI the instructions you wrote off the top of your head and asking it to "rewrite it into a clearer and more structured prompt." AI will supplement any missing conditions.
Conclusion: The Key to Conveying High-Quality Service to AI
The key to conveying high-quality service to AI is to integrate the four elements of role assignment, specific evaluation criteria, sharing ideal examples, and step-by-step feedback into a single prompt. By combining the seven tips and templates introduced in this article, and ultimately having humans verify facts and check for uniqueness, you can elicit high-quality output that accurately reflects the value of your company's services from AI. Along with efforts to enhance the resolution of prompts, it is recommended to inspect your company's communication status using the LLMO Countermeasure Diagnosis Checklist.