95% of People STILL Prompt ChatGPT-5 Wrong

Mastering ChatGPT-5: 5 Tips for Better Results

Key Concepts: ChatGPT-5, model consolidation, invisible router, instruction following, router nudge phrases, verbosity control, prompt optimizer, XML sandwich, perfection loop, second-order effects, meta prompt, text expander app.

GPT-5's Biggest Changes

• Update Number One: Model Consolidation:
  • Previously, ChatGPT Plus users had access to multiple models. Now, there are three: GPT-5, GPT-5 Thinking Mini, and GPT-5 Thinking.
  • OpenAI implemented an "invisible router" to select the appropriate model for each request.
  • The router's performance is inconsistent, leading to varying output quality.
  • OpenAI may prioritize faster, less capable models due to cost considerations.
• Update Number Two: GPT-5 is Much Better at Following Instructions:
  • GPT-5 is designed for AI agents and excels at precise instruction following.
  • It struggles with vague or poorly constructed prompts compared to previous models.
  • Old prompting techniques may yield worse results.

5 Tips to Improve ChatGPT-5 Outputs

Tip Number One: Router Nudge Phrases (Effort: Low)

• Adding specific phrases to prompts can influence the invisible router to select a higher reasoning model.
• Method: Append phrases like "think hard about this" to the end of the prompt.
• Example: Asking "What are the pros and cons of putting my cash in a low-cost index fund versus a money market account?" with and without the nudge phrase. The nudge phrase resulted in a more comprehensive answer including second-order effects.
• Second-order effects: Deeper reasoning outputs often include second-order effects that were not initially considered.
• Pro Tip: Effective phrases include: "Think hard about this," "Think deeply about this," and "Think carefully in contrast with this." Phrases like "This is critical" and "This is very important" were ineffective.
• Pro Tip Number Two: Plus users with access to the thinking model will still benefit from this tip, though free users will see the biggest improvements.

Tip Number Two: Verbosity Control (Effort: Low)

• Specific phrases can control the length and detail of GPT-5's outputs.
• Low Verbosity: Use when only critical information is needed.
  • Example: Drafting a Slack message to a CMO. Phrase: "Give me the bottom line in 100 words or less. Use markdown for clarity and structure."
• Meeting Verbosity: Use when key takeaways and context are required.
  • Example: Explaining click-through rate drops and conversion rate increases in a team meeting. Phrase: "Aim for concise 3 to 5 paragraph explanation."
• High Verbosity: Use for comprehensive documents like project briefs or research summaries.
  • Example: Generating a project brief for an Apple AI strategy overhaul. Phrase: "Provide a comprehensive and detailed breakdown, 600 to 800 words."
• Specific word counts are handled more effectively by GPT-5 than previous models.

Tip Number Three: Use OpenAI's Prompt Optimizer (Effort: Medium)

• OpenAI has an official prompt optimizer tool that rewrites prompts for GPT-5.
• Process: Paste the prompt into the tool, click optimize, and review the changes and rationale.
• Improvements:
  • Adds structure by breaking down text into logical sections.
  • Eliminates vagueness by explicitly stating requirements.
  • Adds error handling by prompting for clarification on contradictions or missing information.
• Free Workaround (Meta Prompt):
  • Use the following meta prompt with ChatGPT-5: "You are an expert prompt engineer specializing in creating prompts for AI language models, particularly ChatGPT-5 thinking model. You're tasked to take my prompt and make it better blah blah blah. Here's my initial prompt" and paste your initial prompt.
  • GPT-5 excels at critiquing and improving its own instructions.
• Pro Tip: Add the meta prompt to a text expander app for quick access.

Tip Number Four: Create an XML Sandwich (Effort: Medium)

• Organize prompts using XML tags to explicitly label different components.
• Concept: Structure prompts like <context>background information</context>, <task>do this thing</task>.
• Example: Instead of a wall of text for a product manager interview, use:
  • <task>Act as a hiring manager and based on my resume and job description, ask me three questions I'm likely to face.</task>
  • <resume>[Paste Resume Here]</resume>
  • <job_description>[Paste Job Description Here]</job_description>
• Pro Tip: Save a template in a text expander with defaults (e.g., <tone>user-friendly and conversational</tone>).
• Pro Tip Number Two: This structure works especially well for custom instructions in custom GPTs and ChatGPT projects.

Tip Number Five: The Perfection Loop (Effort: High)

• Exploit GPT-5's ability to critique itself by instructing it to create its own definition of excellence, grade its own work, and iterate until it achieves the best result.
• Process:
  • Tell GPT-5 to develop an internal rubric for quality.
  • Instruct it to iterate and refine its output until it scores top marks against the rubric.
• Examples:
  • "Write a market analysis report on the enterprise AI industry. Before you begin, develop an internal rubric for what constitutes a world-class market analysis report. Internally iterate and refine the draft until it scores top marks against your rubric."
  • "Draft an outline for my QBR (quarterly business review) presentation. Before you begin, create an internal rubric with five criteria for a perfect QBR. Then use that rubric to internally iterate the outline until your response scores 10 out of 10."
• Pro Tip: Use a universal perfection loop prompt (provided in the video description) to avoid writing custom iteration instructions each time.
• The perfection loop is best for complex zero-to-one tasks like creating finished documents or writing production-ready code.

Conclusion

The five tips presented are not mutually exclusive and can be combined to improve ChatGPT-5 outputs. Nudge phrases, verbosity control, the prompt optimizer, XML sandwiches, and the perfection loop can be used together to achieve optimal results. The perfection loop is particularly useful for complex tasks requiring high-quality outputs.