Do You Have to Be Polite to AI?

Good morning.
Most of us do it without thinking. We type a request into ChatGPT, add a "please," and hit enter. Then we catch ourselves wondering why. About 70% of AI users do exactly that. This week's post digs into whether it actually matters — and finds a more useful habit buried underneath the question.

First time reading? Get your own free subscription here.

AI INSIGHT
Do You Have to Be Polite to AI?

I type "please" into ChatGPT and Claude more than I would like to admit. Not every time. But often enough that I have started noticing it. I will fire off a request, add "please" at the end, hit enter, and then sit there for a second wondering why I just did that.

Turns out I am in very good company.

Around 70% of AI users regularly say "please" and "thank you" to chatbots, according to surveys conducted across the US and UK. Most say they do it because it is simply the right thing to do, whether they are talking to a person or a machine. A smaller group admits something more interesting: they are being nice now, just in case.

That last group has a theory, at least. The majority are running on autopilot, the same way they hold a door for strangers. We have been practicing politeness since we were four years old. Asking us to turn it off when we open a chat window is like asking us not to wave back at someone who waves first.

Stanford researchers documented this pattern back in 1996, decades before ChatGPT existed. People were already being polite to computers, even hesitating to criticize a machine when the machine itself was asking for feedback. Our brains follow social scripts. They do not always check whether the audience is human first.

But here is the part nobody talks about.

Politeness actually works.

Research has found that polite prompts can improve AI accuracy by roughly 9%. Not because the AI is grateful. Because people who phrase requests politely tend to write more carefully. They add context. They explain what they actually want. The courtesy is almost beside the point. The clarity is what matters.

A January 2026 peer-reviewed study reinforced this finding: users who spoke to AI in a more conversational, natural way framed their requests more thoughtfully. Better tone, better communication, better output. Not because the model has feelings. Because decent communication and useful communication tend to travel together.

None of this means you need to pepper every prompt with "if it is not too much trouble." Sam Altman, CEO of OpenAI, has noted publicly that the extra polite phrases add up across billions of daily requests. Every word you type runs through the model's processing. "Thank you" is not free, in the most literal sense.

There is also something more useful you could be doing instead.

A recent Anthropic analysis found that only about 30% of users give their AI tools any instruction about how to respond. Most people say please. Far fewer say things like:

That second kind of instruction does more for your results than any number of pleases ever will. And if you want to see what people actually want most from AI, it turns out clarity and control rank higher than almost anything else.

Want to try this yourself?

Paste this into ChatGPT, Claude, or any AI tool you use:

Ask that once a week for a month. You will learn more about getting results from AI than any tutorial can teach you.

So do you have to be polite?

No. But the instinct to explain yourself rather than demand, to give context rather than bark orders, is the best communication habit you can bring to any AI conversation.

Turns out please was never magic. But then, it never really was.

WHERE TO GO NEXT
If typing out clearer prompts feels like friction, try speaking them instead. Wispr Flow lets you dictate into any text field naturally — and natural speech tends to be exactly the kind of clear, contextual prompting this article is talking about.

If the prompt-writing angle grabbed you, this is the natural next read: How to Use ChatGPT to Spark (Not Steal) Creative Ideas