Why Being Polite to AI Might Be Costing More Than You Think
When someone says “please” or “thank you,” it usually doesn’t seem like a big deal. It’s just basic manners, right? But in the AI world, those extra words might be doing more than we think, and not just socially.
There’s a post that’s been making the rounds on LinkedIn, and it’s got people talking. It points out that all those polite prompts, such as “hi, hope you’re well” and “thanks in advance,” might be costing us more than we realize—not emotionally, literally.
We’re talking about increased GPU time, higher electricity usage, and a growing bill, all because users are being polite to chatbots.
It sounds a bit over the top, but it’s not wrong.
Being nice has a cost
Here’s what’s happening behind the scenes: every word you type into an AI model takes compute power to process. Longer prompts mean more data. More data means more tokens. More tokens? More energy. Multiply that across millions of users and it starts adding up.
And it’s not only about cost; it’s also about speed, server load, and even environmental impact. In other words, every polite prompt is less efficient than getting straight to the point.
So why do we still do it?
Because we’re human.
We’re wired to be polite when we’re talking to something, even if that something is a machine, especially when the machine talks back like a person. The more lifelike these systems sound, the harder it is for us to treat them like tools.
We say “please” and “thank you” because it feels weird not to. After all, it feels right, and its instinct.
That instinct is good, but it’s also what’s causing this polite-prompt overload in the first place.
What does the model actually learn from this?
Here’s where it gets interesting: the way we speak to AI shapes how it responds.
When we fine-tune these models using polite language, they start to expect that tone. They associate softer phrasing with certain types of requests or behavior. And that influences the answers we get back.
It’s not just us who are training the model to be helpful. Our tone reflects on us. Some have even claimed that AI responds much better when we are polite.
And yeah, it goes both ways
Suppose we adjust how we speak to AI, and AI adjusts how it responds based on that. We’ve got a feedback loop. It’s about behavior, influence, and culture now, not just efficiency or performance.
In trying to make AI sound more human, we’ve started treating it like a human. And that’s changing how we interact with tools in general.
So… should we stop being polite?
Not necessarily. It’s not about turning into robots ourselves. But we should think more critically about the design choices we’re making and how they affect behavior at scale.
Do we want AI tools to feel conversational? Sure. But do we want to train people to treat machines like they have feelings? Probably not.
When we spend millions just to accommodate “thanks so much,” we need to ask whether we are optimizing for humanity or just for comfort.
Final thought
This whole conversation isn’t really about being polite. It’s about being more intentional.
Polite prompts might use a little more energy but so do all the vague ones. Think about how often we type things like “rewrite this,” “make it better,” or “can you fix this?” without giving much direction.
That back-and-forth eats up way more compute than a simple thank you ever will.
So, the real issue isn’t whether we’re too polite to AI. Perhaps we need to get better at saying what we actually mean.
Because clearer prompts mean fewer do-overs. And fewer do-overs mean smarter, faster, more useful AI for everyone and less consumption.
