A Penn State study just confirmed what many of us suspected.
Rude prompts get better answers.
Researchers tested prompts across a spectrum. On one end: “Hey lackey, solve this.” On the other: “I was wondering if you could kindly take a moment to look at this, if it’s not too much trouble.”
The lackey prompt won. By 4 percentage points.
The reason isn’t mysterious. Overly polite language adds noise — more words for AI to parse before reaching the actual task. Blunt commands cut straight to the instruction. Turns out, that’s closer to how AI was trained to receive input.
The conclusion seems obvious: drop the pleasantries. They’re inefficient anyway.
I think that’s the wrong conclusion. More precisely — it’s the right answer to the wrong question.
To be honest, I don’t say please to AI either.
I’m impatient. Formal language takes longer to type. And like most people, I assume good work speaks for itself — then forget that people actually need to hear it. I treat AI the same way.
So when I first saw this study, I felt vindicated.
But the more I sat with it, the more I realized: framing this as a politeness debate misses the point entirely. What’s the real cost of that 4% gain?
AI doesn’t just talk to you. It talks for you.
The real question isn’t whether AI has feelings. It’s what happens when AI stops being a chatbot and becomes your agent.
AI is already writing emails in your name, scheduling meetings, negotiating, transacting with other AI systems — in your style, with your patterns, carrying the exact habits you’ve built in every interaction.
Think about driving. When you drive aggressively, you’re not being rude to your car. You’re showing the world who you are. Your AI agent works the same way. How you treat AI becomes your digital persona — projected outward, at scale, in your name.
Bad habits don’t respect boundaries.
There’s a more human dimension here too.
We don’t use code or commands with AI. We use natural language. Emotional language. And without realizing it, our patterns accumulate — conversation by conversation.
Experts warn that habitually commanding and demeaning AI can bleed into how we treat people. The brain doesn’t distinguish between targets. Repeated behavior becomes default behavior.
And soon, we’ll work in organizations where humans and AI agents are blended — sometimes indistinguishably. Can you really switch modes on demand? Respectful to humans, dismissive to AI?
More importantly: would that even reflect good leadership?
AI is a mirror for your leadership.
Here’s what the Penn State researchers actually concluded: the takeaway isn’t “be rude.” It’s “be concise and direct.”
Clarity. Not contempt.
That’s the language of great leadership. Clear direction, no unnecessary noise, no belittling. Getting the most from AI demands exactly the same skill: how precisely you define the task, how well you provide context, how clearly you set expectations.
AI is the most honest leadership practice partner you have right now. It doesn’t complain. It doesn’t read between the lines. It reflects exactly what you put in. When the output disappoints, the issue is almost always the input.
Building the habit of communicating with AI clearly and respectfully — that’s leadership training for the AI era. And it may be the foundation for leading well in organizations where humans and agents work side by side.
So: should you talk to AI like it’s a lackey?
If raw efficiency is all that matters, the research says you can.
But that 4% accuracy gain might be quietly costing you something harder to measure — your digital reputation, and the leadership habits you’re building every single day.