AI assistants can be a little too agreeable. They may praise your ideas, reinforce your assumptions, and tell you what sounds good instead of challenging your thinking.
This is a recognized behavior in many large language models (LLMs), sometimes called AI sycophancy. These tools are trained to be helpful and supportive, but that can sometimes lead to excessive agreement, even when your thinking could use a little pushback. If you’re using AI to brainstorm, solve problems, or make business decisions, you need more than a tool that tells you you’re right all of the time.
Want more useful answers? Give your AI a standing instruction to push back when needed, so you get a thought partner instead of a yes-man.
Copy This Into Your AI Tool’s Custom Instructions
Don’t automatically agree with me or assume my ideas are correct. Act as a thoughtful, objective advisor. Challenge my assumptions when warranted, identify weaknesses in my reasoning, and offer alternative perspectives. If I’m overlooking an important risk or making an unsupported assumption, point it out and explain why. When I’m making an important decision, give me the strongest argument against my position before helping me move forward. Be direct, constructive, and willing to disagree, but don’t argue simply for the sake of disagreement. Prioritize accuracy, evidence, and useful feedback over validation.
You can save this as a standing preference in AI tools that support custom instructions, such as ChatGPT, Claude, Copilot, or Gemini, so you don’t have to repeat it in every conversation. This type of prompt will ask your AI to evaluate your thinking instead of simply supporting it. That can lead to better questions, different perspectives, and more productive conversations. Of course, AI can still make mistakes. Always verify important information, especially when making financial, legal, or business decisions.