Getting Better Results from ChatGPT

ChatGPT's Steerability

GPT models are highly steerable—they respond precisely to instructions. This means your prompts have significant control over the output style, length, format, and approach.

Use this to your advantage by being explicit about what you want.

Clean Up Contradictory Instructions

Contradictory or conflicting instructions hurt performance. Review your prompts for conflicts:

Problematic: "Be concise but also thorough and detailed."

Better: "Prioritize the three most important points. For each, give one supporting detail."

Problematic: "Be creative but stick exactly to the template."

Better: "Follow this template structure, but use creative language for the descriptions."

If you're getting inconsistent results, check your prompt for mixed signals.

Start Simple, Then Expand

ChatGPT works well with iterative refinement. Start with a simple request and build from there:

1. First prompt: "Write a cold email for my consulting services"

2. Refine: "Make it shorter and focus on the pain point of wasted time"

3. Refine again: "Add a specific example of how we helped a similar company"

4. Final touch: "Make the CTA softer—suggest a quick call instead of demanding a meeting"

This approach often produces better results than one complex prompt.

Use Role-Playing Effectively

Assigning ChatGPT a specific role improves results for many tasks:

"You're an experienced copywriter. Improve this headline."
"Act as a marketing strategist reviewing my campaign plan."
"You're a potential customer. What questions would you have about this product?"

Roles help ChatGPT adopt the right expertise and perspective.

Custom Instructions for Persistent Context

Use ChatGPT's custom instructions feature to set context that applies to all conversations:

Your role and industry
Preferred response style (concise vs. detailed)
Formatting preferences
Common tasks you need help with

This saves you from repeating the same context in every conversation.

Be Explicit About Proactive vs. Conservative

Tell ChatGPT whether you want it to take initiative or wait for guidance:

For proactive behavior: "Go ahead and make changes you think would improve this. I'll review afterward."

For conservative behavior: "Don't make any changes yet—just identify issues and suggest fixes for my approval."

Without this guidance, you may get either too much or too little action.

Request Markdown Formatting Explicitly

ChatGPT doesn't always use markdown formatting automatically. If you want headers, bold text, or bullet points, request it:

"Format your response with clear headers and bullet points"
"Use markdown formatting for readability"
"Structure this with H2 headers for each main section"

Request Structured Output

ChatGPT handles structured formats well when asked:

"Create a table comparing these three options"
"Output this as a JSON object with fields: name, description, price"
"Format as a numbered checklist"
"Give me this as markdown I can paste into a document"

Save Effective Prompts

When you find a prompt that works well, save it:

Keep a document of prompts that produced great results
Note what made them effective
Adapt them for similar future tasks

Reusing proven prompts saves time and maintains quality.

Handle Verbose Output

If ChatGPT is too wordy, explicitly request brevity:

"Keep your response under 100 words"
"Answer in 2-3 sentences max"
"Be concise—skip the preamble and give me the answer directly"

ChatGPT respects length constraints when you specify them.