95% of People STILL Prompt ChatGPT-5 Wrong

Understanding the Changes in GPT-5

Introduction to User Experience with GPT-5

• After the launch of ChatGPT-5, many users reported a decline in output quality despite unchanged prompting techniques.

• The core issue lies in OpenAI's fundamental changes to GPT-5's architecture, which affects how previous prompts perform.

Key Updates in GPT-5

Model Consolidation

• OpenAI has consolidated models from multiple options to just three: GPT-5, GPT-5 Thinking Mini, and GPT-5 Thinking.

• This consolidation introduced an "invisible router" that determines which model processes user requests, leading to inconsistent results based on prompt submission.

Improved Instruction Following

• While GPT-5 excels at following explicit instructions due to its training for AI agents, it struggles with vague or poorly constructed prompts.

Tips for Optimizing Outputs with GPT-5

Tip 1: Router Nudge Phrases

• Adding specific phrases at the end of prompts can influence the invisible router to select a higher reasoning model.

• Example: Using "think hard about this" can trigger deeper reasoning compared to vague terms like "important."

Pro Tip for Effective Phrasing

• Three effective nudge phrases identified are:

• "Think hard about this"

• "Think deeply about this"

• "Think carefully"

Additional Resources for Mastering Prompt Engineering

HubSpot’s Free Playbook Recommendation

• For those serious about mastering these techniques, HubSpot offers a free playbook on prompt engineering that emphasizes building systems over one-off prompts.

Verbosity Control Techniques

Tip 2: Controlling Output Length

• Users can control output length by using specific phrases that adjust verbosity settings within the invisible router.

Understanding AI Prompt Optimization

The Importance of Context in Team Communication

• Teams require more than just numerical data; they need context to understand the implications and actions required.

• A concise explanation, ideally 3 to 5 paragraphs, can effectively convey necessary details without overwhelming the audience.

• High verbosity outputs are beneficial for comprehensive documents like project briefs or research summaries.

Project Brief Example: Overhauling AI Strategy

• An example project involves a complete overhaul of Apple's AI strategy, emphasizing the need for detailed breakdowns (600 to 800 words).

• GPT-5 is noted for its improved handling of specific word counts compared to previous models.

Utilizing OpenAI's Prompt Optimizer Tool

• Many users are unaware that OpenAI offers a prompt optimizer tool that enhances prompts for GPT-5.

• The tool provides optimized prompts and allows users to review changes, helping them learn how to write better prompts themselves.

Key Improvements from the Optimizer Tool

• The optimizer consistently adds structure by breaking down complex prompts into distinct sections.

• It eliminates vagueness by ensuring all arguments are based on provided achievements.

• Error handling is enhanced as it reminds users to clarify contradictions or missing information in their prompts.

Free Workaround for Prompt Optimization

• A free alternative involves using a meta prompt with ChatGPT, allowing users to improve their initial prompts effectively.

• This method leverages GPT-5’s ability to critique and enhance its own instructions.

Structuring Prompts with XML Tags

Creating an XML Sandwich

• Using XML tags helps organize instructions clearly, which is crucial given GPT-5's precision in following structured commands.

• Instead of presenting unstructured text, labeling components (e.g., background information, task description, output format) improves comprehension and outcomes.

Practical Application of Structured Prompts

• For instance, instead of dumping all information at once when preparing for an interview, one should clearly define tasks and provide relevant background separately.

The Perfection Loop: Iterative Improvement

Leveraging Self-Critique in GPT Models

• OpenAI suggests utilizing GPT's self-critiquing capabilities by instructing it to create its own definition of excellence before generating responses.

Examples of Iterative Processes

1. Market Analysis Report:

• Before drafting, develop an internal rubric for what constitutes a high-quality report and refine until it meets top standards.

2. Quarterly Business Review Presentation:

• Create criteria for a perfect QBR outline and iterate until achieving optimal results.

Universal Perfection Loop and Its Applications

Understanding the Universal Perfection Loop

• The universal perfection loop is a method that allows pacing to the end of any prompt, facilitating structured completion of tasks.

• It is particularly effective for complex "zero to one" tasks, such as creating finished documents from scratch or writing production-ready code.