ChatGPT and Gemini Give Better Responses With This Simple Custom Instruction

Artificial intelligence has become a daily tool for writing, coding, brainstorming, and research. 

But many users notice one common issue with AI chatbots like ChatGPT and Gemini—they often agree with almost everything you say, even when your assumption is incorrect.

A simple change in the custom instructions can encourage these AI assistants to evaluate your ideas more carefully before responding. 

While this doesn’t completely change how the models work, it can make conversations more useful by reducing unnecessary agreement.

Why AI Sometimes Agrees Too Easily

Large language models are generally designed to be helpful and cooperative. 

As a result, they may focus on answering your question instead of checking whether your original assumption is accurate.

For example, if you ask a question based on incorrect information, the chatbot might continue the discussion instead of pointing out the mistake first. 

This can lead to less reliable responses, especially when you’re making important decisions or solving complex problems.

A Simple Custom Instruction That Can Help

You can improve the quality of responses by adding a custom instruction that tells the AI to challenge incorrect assumptions.

A prompt like this works well:

“If I make a factual mistake or a weak assumption, point it out before answering. Challenge incorrect reasoning instead of agreeing with it. Don’t praise ideas unnecessarily, and don’t invent disagreements when my reasoning is sound.”

This encourages the chatbot to review your premise before generating an answer.

Where to Add the Instruction

• ChatGPT: Go to Settings → Personalization → Custom Instructions.
• Gemini: Go to Settings → Personal Intelligence → Instructions for Gemini.

Once saved, the instruction applies to future conversations until you change or remove it.

Does It Really Work?

Yes—but with some limitations.

Custom instructions influence how ChatGPT and Gemini respond, but they do not completely change the underlying AI model. 

Sometimes the chatbot will challenge weak assumptions, while other times it may still miss factual errors.

Claude is often praised by users for naturally questioning incorrect claims more frequently without requiring additional instructions. 

However, adding a custom instruction can noticeably improve the experience with ChatGPT and Gemini for many everyday tasks.

Conclusion

If you use ChatGPT or Gemini regularly, spending a minute updating your custom instructions is worth trying. 

While it won’t make either chatbot identical to Claude, it can encourage more thoughtful responses and help catch mistakes before they become bigger problems.