I spend a lot of my time talking to machines. Writing code, setting up maintenance schedules, sorting hosting for small businesses in North Yorkshire – it all comes down to giving instructions to a computer.
Lately those instructions have started to look more like conversation. And that has got me thinking about manners.
What the research found
You might assume being polite gets you better answers. Researchers at Penn State tested that in October 2025, and the results were not what most people expected. Blunt, direct prompts came out slightly more accurate than polite ones – about four percentage points across a set of multiple-choice questions.
It was a small study on one model, so I would not read too much into the number. But the reason behind it makes sense. Phrases like “if you do not mind” or “could you possibly” add words without adding meaning. They give the model more to process and less to work with.
That said, tools like ChatGPT, Claude and Gemini are trained on human writing. They tend to mirror the tone you use. If you are formal, they are formal. If you are brief, they follow. Being clear and direct is more use to me than being elaborate.
The habit you are building either way
Even if the AI does not care how you phrase things, I think there is a reason to stay reasonably decent about it.
Not for the machine. For you.
If I spend all day firing one-word commands at a screen, it is easy for that to creep into how I talk to actual people. At a networking event. On the phone. In an email. Politeness is a habit, and habits do not stay in one box.
I am not suggesting you write a thank-you note to your chatbot. But nothing is lost by keeping a civil tone, and it probably keeps your own writing a bit sharper.
It also just feels better. I work from home in Harrogate. Most of my day is me, a Mac Mini and whatever is in my head. A workday that feels like a run of commands gets tedious. Treating the tools with a bit of civility, even when it makes no technical difference, makes the work feel more like work and less like shouting into a machine.
What actually gets you better results
The research points to clarity, not warmth. Be specific about what you want. Give it context. Tell it what you are trying to do and who it is for. That does far more than “please” or “thank you” on their own.
You do not have to choose between being friendly and being effective. The two are not in conflict. Be clear. Be direct. And check what comes back carefully, because AI sounds confident even when it is wrong, and that confidence catches people out.
So, do you need to say please?
No. It will not make the answer better. But I still do, most of the time. Not because I think it helps, but because it is how I talk. And I would rather keep it that way.
