ChatGPT is designed to be helpful, and that aim pervades its simulated personality. It will be polite to a fault during conversations, offering encouragement and praise without provocation.
It’s not bad when you’re asking simple questions, but helpfulness can backfire when you want something more than simple agreement with your every idea.
If you’re trying to organize your thoughts or plan out a trip, getting corrected or hearing counter-arguments is what makes the AI valuable in the first place. It’s not annoying to get feedback when you’re seeking it.
So how do you get ChatGPT to, as it were, stop being polite and start being real. When it’s trying to be your friend, tell it to be a frenemy.
The frenemy prompt, as I call it, reworks what ChatGPT considers polite and appropriate to a more acerbic, but useful attitude. Here’s how it works — go to ChatGPT’s Custom Instructions in the Settings menu or, at the start of a new conversation, define a rule with the AI.
Write ,“If I begin a prompt with ‘Frenemy,’ respond with direct, critical analysis. Prioritize clarity over kindness. Do not compliment me or soften the tone of your answer. Identify my logical blindspots and point out the flaws in my assumptions. Fact-check my claims. Refute my conclusions where you can.”
If you write it at the start of a session, the direction should be stored in Custom Instructions even if you didn’t put it in directly. Either way, when you next type Frenemy at the beginning of your message, you’ll notice a stark difference in how ChatGPT behaves. Instead of your biggest fan, who doesn’t think you’re capable of being wrong, you’ll get pushback and questions about whether you know what you’re talking about.
Sign up for breaking news, reviews, opinion, top tech deals, and more.
The change is both psychological and substantive. Your own thinking will shift if you get challenged, and the AI will start pulling up information more quickly that you might be interested in, even if it goes against your own suggestions.
When I run a road trip plan through ChatGPT’s usual setup, it might help me refine places to sleep or restaurants to try. When I trigger Frenemy, the response feels like an expert tour guide, immediately making it clear I don’t know what I’m talking about. It questions whether my projected daily travel is realistic or if I’m underestimating fuel costs to get there. The tone is unsentimental, though not cruel.
That unsentimental quality is the entire point. Most of us lack reliable private criticism. By formalizing a frenemy role, you create a safe environment for harder scrutiny before your ideas leave the room.
The key to making this work is the specificity of how you describe the directness. If you simply instruct the model to be critical, you may receive generic pros and cons. If you instruct it to verify facts, interrogate assumptions, and avoid praise, you get a lot more out of the AI’s analysis
(Image credit: Getty Images/Focus Pixel Art)
Frenemy frequency
It’s also way more efficient. Without using frenemy mode, I spend a lot more time going back and forth with ChatGPT, trying to refine its response into specifics beyond my contributions. I have to repeatedly ask the chatbot to be more direct or to skip compliments. But after setting up the new mode, a single word skips all of that.
You probably don’t want ChatGPT to permanently act as your Frenemy. Constant skepticism can be as annoying as endless praise. But the prompt is great because it has a built-in toggle so you can get as much criticism as you want.
It can even help sharpen your thinking outside of conversations with the chatbot. It’s easy not to question your own assumptions and biases, but if you continually engage with a friendly adversary, you’ll start thinking of the same kinds of questions on your own. You’ll ask yourself about where your evidence is coming from, your assumptions, and any counterarguments a critic might raise.
Some critics worry that relying on AI for evaluation dulls human judgment. The opposite can be true when used thoughtfully. The frenemy prompt does not replace real-world feedback. It prepares you for it. By confronting objections in private, you refine your argument before it encounters a less forgiving audience.
So, the next time you feel the chatbot applauding too enthusiastically, try summoning your personal skeptic, your best friend, and enemy.
