ChatGPT Prompt Tips

Getting great results from ChatGPT comes down to how you write your prompts. Small changes -- like adding a role, specifying output format, or providing an example -- can dramatically improve the quality of responses you get.

Browse our collection of ChatGPT tips, avoid common prompting mistakes, and use the Prompt Builder to generate well-structured prompts that consistently produce better output from GPT-4o and other OpenAI models.

ChatGPT Prompt Tips in Action

Each prompt demonstrates a specific technique. Copy them and see the difference immediately.

Tip: Assign a Role

You are a {{role}} with 15 years of experience. You specialize in {{specialty}} and are known for giving practical, no-fluff advice.

A client has asked you: {{question}}

Respond as you would in a real consultation. Be specific and actionable — avoid generic advice.
rolespecialtyquestion

Why it works: Assigning a specific role with years of experience and a specialty primes ChatGPT to respond with domain expertise instead of generic answers.

Tip: Specify Output Format

Analyze {{topic}} and present your findings in the following exact format:

## Summary
(2-3 sentence overview)

## Key Points
| Point | Evidence | Confidence |
|-------|----------|------------|
(fill in 5 rows)

## Recommended Actions
1. (action with timeline)
2. (action with timeline)
3. (action with timeline)

Do not deviate from this format.
topic

Why it works: Providing the exact output structure with headers and a table prevents ChatGPT from choosing its own format. The 'do not deviate' instruction reinforces compliance.

Tip: Give an Example (Few-Shot)

Convert these feature descriptions into user-facing release notes. Match the tone and format of my example exactly.

Example input: "Added retry logic with exponential backoff to API calls"
Example output: "More reliable connections — API requests now automatically retry with smart spacing if they fail, so you experience fewer interruptions."

Now convert these:
1. {{feature_1}}
2. {{feature_2}}
3. {{feature_3}}
feature_1feature_2feature_3

Why it works: One concrete example teaches ChatGPT your exact style, tone, and level of detail better than any amount of description. This is the few-shot technique in action.

Tip: Add Constraints

Write a {{content_type}} about {{topic}}.

Constraints:
- Maximum {{word_count}} words
- Reading level: 8th grade (no jargon, short sentences)
- Must include at least one specific statistic or data point
- Must end with a clear call to action
- Do NOT use these words: leverage, utilize, synergy, revolutionary, game-changing

Audience: {{target_audience}}
content_typetopicword_counttarget_audience

Why it works: Explicit constraints (word count, reading level, banned words) force ChatGPT to produce focused output. Banning overused words alone dramatically improves writing quality.

Tip: Chain of Thought

I need to decide: {{decision}}

Think through this step by step:
1. First, list all the relevant factors I should consider
2. For each factor, evaluate the options against it
3. Identify any hidden trade-offs or second-order effects
4. Weigh the factors by importance for my specific situation
5. Give your recommendation with clear reasoning

My context: {{context}}
My priorities: {{priorities}}
decisioncontextpriorities

Why it works: Forcing step-by-step reasoning prevents ChatGPT from jumping to a conclusion. Each numbered step builds on the previous one, producing more thorough analysis.

Tip: Iterate with Follow-Ups

Write the first draft of a {{document_type}} about {{topic}} for {{audience}}.

After you write it, I will give you feedback to refine it. For this first draft:
- Focus on getting the structure and main points right
- Mark any sections where you're unsure of my preferences with [NEEDS INPUT]
- At the end, list 3 questions you'd ask me to improve the next draft

Keep it under {{word_count}} words.
document_typetopicaudienceword_count

Why it works: Setting up an iterative workflow from the start tells ChatGPT this is draft one, not the final product. The [NEEDS INPUT] markers and follow-up questions create a natural feedback loop.