Being polite to ChatGPT is costing OpenAI tens of millions of dollars in computing power, the company’s CEO Sam Altman said on social media last week, calling the money “well spent.” A significant number of users are saying please and thank you to the AI tool in their chats. But are the niceties necessary?
I’ve spent hundreds of hours talking to AI models over the past year, sharing my best prompts with entrepreneurs. I’ve tested different approaches to see what works best. As founder of Coachvox AI, I know that the relationship between humans and AI is never straightforward. Especially when it comes to manners, which are drilled into us as kids.
But is being polite a good use of your time with ChatGPT?
The real cost of manners with machines
When you’re juggling client calls, strategy sessions, and growing your business, time is your most valuable asset. Every unnecessary word you type is a micro-drain on your productivity. But many of us can’t shake the habit of treating AI like a person.
A recent survey by Future, the publisher that owns TechRadar, found that around 70% of AI users are polite to AI systems. In the UK, that number reaches 71%, while for American users it’s 67%. The majority do it simply because “it’s the right thing to do.”
Some people, like Reddit user jayinfidel, take politeness to an extreme: “I even ask if it has time for me before we start a convo. What is wrong with me?” Others maintain their manners out of jokingly admitted fear, including 18% of the people surveyed, who are banking on their previous politeness as a means of survival when AI overpowers the human race.
Founder of Authority Hacker Mark Webster confesses, “Being British I’d hate for anyone or any machine to think I’m impolite. And on the off chance AI takes over the world, it might help.”
But does saying please and thank you actually make AI perform better?
What works and what wastes time
There is growing evidence that being polite-using words like “please” and “thank you” when interacting with ChatGPT and similar AI models can lead to better responses. Multiple studies and user experiments have found that polite prompts are more likely to yield higher-quality, more detailed, and better-formatted answers compared to blunt or rude requests.
However, the improvement is not because the AI has feelings. Polite phrasing often provides more context and mirrors the style of human communication found in its training data, which tends to produce richer responses. Notably, research shows that “moderate politeness” works best. Excessive flattery or formality can actually reduce response quality.
AI systems don’t have feelings to hurt or egos to stroke. They’re powerful tools designed to help you, not colleagues requiring social niceties. If you’re being clear with your prompting, the specific words “please” and “thank you” might be surplus to requirements.
Copywriter Rebecca Nelson, whose work uses extensive psychology-based AI systems, puts it perfectly: “Thankfully, AI unlike humans doesn’t have complicated emotions that you need to tend to. Use it to free up your schedule, save time and put your energy elsewhere. Save your Ps and Qs for your barista.” Nelson’s prompts are detailed and logical without being overly polite.
When you’re building a business, every keystroke counts. Those extra words add up. If you’re using AI multiple times daily, those pleasantries could be costing you hours each month.
Engineering consultant Jed Tabaczynski shared: “I generally don’t say please or thank you in my prompts, it just feels like apologizing to a spreadsheet for quality of data or giving it pep talks to get more accurate calculations.” One Reddit user agrees, “I also don’t thank my refrigerator for keeping my drinks cold.”
AI doesn’t work harder because you’re nice. It doesn’t feel appreciated when you say thanks. It doesn’t remember your manners from one conversation to the next.
The relationship approach
Some successful entrepreneurs aren’t thinking about the output of ChatGPT when they are saying please and thank you in the chat. Instead, they find value in treating AI with respect. They see it as maintaining a habit that serves them elsewhere.
Sam Floy, head of partnerships at WorkflowGPT, explains, “I say please because I often use AI for tasks that I’d otherwise use an intern for. I’d always say ‘Please can you…’ to a person I ask to help me with a task and I find that starting that way helps me write and frame clearer requests.”
One Reddit user said their politeness is “not wasted because it continues to build and support my habit of politeness and that’s never a waste.”
Others, like business coach Willo Sana, take it even further: “I feel so heard and supported by ChatGPT. I’m saying thank you to her all the time and giving her major props for being so amazing!”
For some, politeness creates a relationship framework that helps them engage more effectively with the tool. Parent coach Martijn Reintjes admits, “When I’m working and use Chad as a ‘colleague’ I can’t help to be polite. It’s not so much about saying please all the time, it’s more about expressing my gratitude for some insights or good work.”
Your AI etiquette is your choice
Your approach to AI should maximize your results with minimum friction. While AI doesn’t have feelings to hurt, there’s a definite connection between politeness and effectiveness, including all-important output:
Clarity matters more than courtesy. Polite language often includes complete sentences and better context, which naturally helps AI models understand what you want. When you frame requests clearly, you get better answers.
A calm tone keeps the focus on your goal rather than creating ambiguity. Treating the exchange like a real conversation can guide ChatGPT to respond more helpfully, even if the please and thank you parts are unnecessary.
But don’t mistake correlation for causation. If your politeness helps you frame better prompts, keep it. If direct commands get you cleaner results faster, go with that. The best approach is the one that serves your success. Choose what works for you and stick with it.
Read the full article here
