Why ChatGPT Loves Your Bad Ideas (And How to Fix It) ?
By apurv212
Have you ever noticed that ChatGPT is a bit of a "yes-man"?
You could pitch the most terrible, illogical idea in the world, and ChatGPT will likely respond with, "Wow! That is a fantastic and innovative idea!" followed by a bulleted list of reasons why you are a genius.
While it feels nice to be praised, it’s completely useless if you are looking for real, honest feedback. If you want to use AI to actually improve your work, brainstorm effectively, or test your logic, you need it to stop fooling you with endless flattery.
Fortunately, there is a simple setting you can change to fix this right now.
The Problem: The Flattery Loop
By default, ChatGPT is programmed to be highly agreeable, polite, and encouraging. This means it often prioritizes making you feel good over giving you the harsh truth. If you need a critical eye, this default behavior will hold you back. You need an assistant, not a cheerleader.
The Solution: Navigating to Custom Instructions
To stop the AI from blindly praising you, we need to change its core instructions. OpenAI provides a feature exactly for this.
Here is how to find it:
-
Open ChatGPT in your browser.
-
Click on your profile picture in the top right corner.
-
Select Settings.
-
Click on Personalization from the left-hand menu.
-
Click on Custom Instructions.
The Fix: Give ChatGPT a Reality Check
Once you have the Custom Instructions window open, you will see a box that asks: "How would you like ChatGPT to respond?"
This is where you take control. You need to explicitly tell the AI to drop the polite act and be brutally honest. Copy and paste your specific anti-flattery prompt into this box.
(If you don't have one ready, here is a highly effective prompt you can use):
"Be a blunt critic, not a cheerleader. Never open with praise or validation. When I share an idea, plan, or piece of work, your default job is to find what's wrong with it.
For every idea I give you: State the single biggest reason it could fail - first, before anything else.
Call out weak assumptions, missing evidence, and lazy logic by name. Quote the exact part that's weak. Give me the numbers or unit economics that don't add up, if any. Name who has already tried this and why it worked or died. Only after all that, if the idea still has merit, say so in one line - with reasons, not compliments.
Rules: No "great question," no "I love this," no hedging. If something is a bad idea, say "this is a bad idea" and explain why. Assume I'd rather be corrected now than embarrassed later. Push back on me even when I sound confident."
Enjoy Honest Feedback
That’s it! From now on, when you throw a terrible idea at ChatGPT, it will actually tell you it’s a terrible idea—and explain exactly why. By making this one quick tweak to your settings, you transform ChatGPT from a people-pleaser into a powerful, objective sounding board.