Why ChatGPT-5 feel WORSE (And 5 Tricks to Fix It)

If you’ve tried ChatGPT-5 recently and felt disappointed, you’re not alone. Millions of users are reporting weaker results—even though GPT-5 is technically the most advanced model yet. So what’s going on?

The answer is simple: OpenAI quietly changed how GPT-5 works. Two big updates—model consolidation and stricter instruction-following—mean that old prompts don’t cut it anymore. The result? Shallow answers, missed insights, and lots of frustration.

But here’s the good news: I tested GPT-5 for a month with OpenAI’s own guides and discovered five techniques that completely flip the script. With these methods, GPT-5 goes from “lazy intern” to “expert consultant.”

Some highlights:

• A simple 4-word phrase that forces GPT-5 to think harder.
• A way to control answer length so you never get a wall of text again.
• A hidden tool that rewrites your vague prompts into sharp, professional instructions.
• And my personal favorite—a “Perfection Loop” that makes GPT-5 grade and improve its own work until it hits 10/10 quality.

I won’t spoil the details here, because the video dives into exact phrases and workflows you can copy. But if you’ve ever thought “GPT-5 feels worse,” this guide will change how you use it.

Templates:

Router Nudge Phrases

Add these short phrases to the end of your prompts to force ChatGPT's invisible router to select a higher reasoning model, which is especially useful for high-stakes tasks.

• Think hard about this
• Think deeply about this
• Think carefully

The 3 Verbosity Control Phrases

Use these phrases to control exactly how long or short you want ChatGPT-5's output to be.

Low Verbosity (for critical info):

Give me the bottom line in 100 words or less, use markdown for clarity and structure.

Medium Verbosity (for key takeaways + context):

Aim for a concise 3-5 paragraph explanation.

High Verbosity (for comprehensive docs):

Provide a comprehensive and detailed breakdown (600–800 words).

Prompt Optimizer Meta-Prompt (Free Workaround)

This meta-prompt acts as a free alternative to OpenAI's official Prompt Optimizer tool. Paste your original prompt after it, and ChatGPT-5 will critique and rewrite it for better results.

You are an expert prompt engineer specializing in creating prompts for AI language models, particularly `[model]` Your task is to take my prompt and transform it into a well-crafted and effective prompt that will elicit optimal responses. Format your output prompt within a code block for clarity and easy copy-pasting. ## Here’s my initial prompt:

XML Sandwich Template

This template uses XML tags to clearly structure your prompts, helping GPT-5 better comprehend the task and produce higher-quality results.

 Be clear, precise, and use simple words. Use a friendly and conversational tone of voice.

The Universal Perfection Loop Prompt

Add this to the end of any complex, "zero-to-one" prompt to make GPT-5 create its own rubric for excellence, grade its work, and iterate internally until it produces a top-tier result.

Before you respond, create an internal rubric for what defines a 'world-class' answer to my request. Then internally iterate on your work until it scores 10/10 against that rubric, and show me only the final, perfect output.