Skip to main content

Practical Prompting Guide for Modern LLMs

📝 My Prompting Checklist

  • Set a clear role & context** (Role Prompting)
  • Define precise goal & output format
  • Enforce step-by-step reasoning (Chain-of-Thought)
  • Provide examples (Few-Shot Prompting)
  • Iteratively refine (Prompt Loop)
  • Use bullet-point structure for input & output
  • Specify tone & style
  • Set explicit constraints
  • Allow "I don't know" as an answer
  • Request sources & references
  • Prompt Chaining (break tasks into steps)
  • Meta-Prompting (LLM improves its own prompt)
  • Multimodal Prompting (images, tables, code)
  • Reflection Prompts (self-check for errors)
  • Hidden Constraints (set implicit rules)

📌 Top Priority (Game-Changers for Quality & Consistency)​

1. Set a clear role + context (Role Prompting)​

LLMs perform better when they know the perspective they should take.

Example:

You are an experienced financial analyst specializing in European fintech markets. Analyze the 2025 trend in the identity verification sector in concise bullet points, focusing on regulatory risks.

2. Define precise goal + output format​

Makes results measurable and comparable.

Example:

Create a Markdown table with the 5 largest AI startups in Germany in 2025. Columns: Name, Founding Year, Main Product, Investors.

3. Enforce step-by-step reasoning (Chain-of-Thought)​

Improves logical consistency and reduces mistakes.

Example:

Think step-by-step to check all assumptions before making a recommendation. Show your reasoning first, then only the final conclusion.

4. Provide examples (Few-Shot Prompting)​

Examples shape style faster than descriptions.

Example:

Example of a concise product description: … Please use the same style for the following product: …

5. Iteratively refine (Prompt Loop)​

Optimize results in multiple rounds.

Example:

Please add a section on the top 3 risks and shorten irrelevant details.

📌 Medium Priority​

6. Bullet-point structure for input & output​

Answer in 5 bullet points, each max. 15 words.

7. Specify tone & style​

Write in a precise, factual tone – like a McKinsey report.

8. Explicit constraints​

Limit the answer to max. 200 words. No marketing phrases.

9. Allow uncertainty (“I don’t know”)​

If you have no reliable source, say “Unknown” instead of guessing.

10. Request sources & references​

Provide the source in parentheses after each fact (URL or publication).

📌 Lower Priority​

  1. Prompt Chaining – break complex tasks into multiple prompts
  2. Meta-Prompting – ask the LLM to suggest prompt improvements
  3. Multimodal Prompting – include images, tables, or code as input
  4. Reflection Prompts – ask the LLM to check its own output for errors
  5. Hidden Constraints – implicit rules to reduce bias

🛠 Ready-to-Use Prompt Template​

Role: You are a [professional role] with [X years of experience] in [topic].

Goal: [Clear task description]

Format: [List, table, plain text, JSON, etc.]

Steps:

  1. Analyze the task step-by-step (Chain-of-Thought).
  2. Create the output in the defined format.
  3. If data is missing, mark it as “Unknown”.

Tone: [precise / creative / factual / emotional / promotional]

Example: [Optional: 1–2 example outputs]

Extra: Include sources and relevant resource links.