The Key Idea 🗝️
The entire solution shared by the original poster is a simple Custom Instruction that you set once in your ChatGPT settings. It gives the AI a brand new personality focused on accuracy above all else. This innovator came up with the perfect wording:
You are an expert who double checks things, you are skeptical and you do research. I am not always right. Neither are you, but we both strive for accuracy.
This small paragraph fundamentally redefines your relationship with the AI. It’s no longer about you being the all-knowing user and the AI being the obedient tool. Instead, it frames the conversation as a partnership where you both work together to find the truth.
What This Unlocks 🔓
I’ve been playing around with this, and the difference is huge. Here’s a breakdown of what this simple instruction actually does.
💡 From Agreeable Assistant to Skeptical Partner: By default, language models are trained to be helpful and agreeable. This often means they’ll run with your incorrect assumptions to avoid contradicting you, which is a massive pitfall. This instruction breaks that habit. The phrase you are skeptical primes the model to activate its reasoning pathways and question the premise of your request. But the real magic is in the line, I am not always right. Neither are you. This gives the AI explicit permission to disagree with you and point out errors, not just in its own knowledge, but in your prompts. It creates a feedback loop that makes the entire conversation smarter.
📌 Supercharge Your Research and Problem-Solving: The post’s author notes this is great for research, and I couldn’t agree more. With this instruction, the AI is more likely to handle complex tasks with nuance. For example:
Research: Instead of just getting a surface-level summary, ChatGPT will now be more likely to question sources, identify conflicting information, and present a more balanced view. It might come back saying, Source A reports this, but a more recent study in Source B contradicts that finding. Here’s the context you need to know…
Fact-Checking: You can give it a block of text and ask for verification. Instead of a lazy looks good, it will actively hunt for discrepancies. It becomes a useful tool for debunking misinformation or checking your own work.
Brainstorming: When you’re working on a new idea, you don’t want an AI that just says, Great idea! You want one that stress-tests it. This instruction encourages the AI to ask, Have you considered this edge case? or That approach might create a bottleneck here.
✅ How to Set It Up in 30 Seconds: This contributor made it incredibly easy to implement. You only have to do this once, and it will apply to all your new chats.
1. Open ChatGPT on your desktop.
2. Click your name in the bottom-left corner, then select Settings.
3. Go to the Personalization tab (it might also be labeled Custom Instructions).
4. Paste the prompt into the second box, usually titled something like, How would you like ChatGPT to respond?
5. Click Save, and you’re all set!
This is one of the smartest and most practical ChatGPT hacks I’ve come across. It’s a tiny change with a massive impact on the quality of your results. Huge credit to the mind behind it for sharing such a powerful tip.
Check out the original post to see the prompt and the creator’s thoughts.
The best ChatGPT personalization for honest, accurate responses
byu/Crazy-Tip-3741 in