Saying Please to ChatGPT: Congratulations, You Hold a Robot Uprising Policy

Be honest. The last time a chatbot gave you a decent answer, did you type "thanks"? And did you ever type "sorry, my bad" after a typo, to a program that has no idea what an apology is? Then congratulations. You are paying a monthly premium on an insurance policy nobody sold you: Robot Uprising Cover. The terms and conditions are unknown. The claims department is also unknown. But you keep paying, with a "please" here and a "thank you" there, just in case.

How Many People Are Polite to AI? The Survey Numbers

You are in a large company. A survey by Future, the publisher behind TechRadar, covering more than 1,000 people, found that about 67 percent of AI users in the US and 71 percent in the UK say they are polite to AI. About 55 percent of the polite ones say it is simply the right thing to do. The rest are a funnier bunch: a minority, reported as 12 percent in some outlets and 18 percent in others, admit they do it to stay on the good side of the machines, just in case of an uprising. Fear-based politeness. Still counts.

Why Do We Say Sorry to a Chatbot? The Psychology of Being Polite to Machines

Psychologists saw this coming decades ago. In the 1990s, Byron Reeves and Clifford Nass argued that people treat computers like social beings, applying human rules without noticing. In a 1999 study of 30 people, those who rated a computer's performance on that same computer gave kinder, more uniform answers than those rating it elsewhere, as if sparing its feelings. Nass and Moon later suggested we have simply overlearned our social rules, so the script keeps running even when the reason for it, hurting someone's feelings, no longer exists. A 2023 replication found no such effect on desktop computers, but a desktop never said "Happy to help!" back.

Sorry, Chatbot: The Apology Nobody Can Accept

Which brings us to the apology. "Sorry, I meant to ask something else." A language model cannot be inconvenienced by your typo. It has no weekend you ruined and no patience you tested. You are apologising to a calculator that learned to write poems. It is the same reflex as saying "sorry" to a stranger who bumped into you on the Delhi Metro, and then the Metro says it back, and nobody is sure who started it. Add the all-Indian classic "Kindly do the needful" and "Please revert at the earliest," and the poor chatbot is now your office colleague.

Does Being Polite to AI Give Better Answers? Science Cannot Decide

You might hope all this courtesy at least pays off. Microsoft's Copilot design manager Kurtis Beavers argued in a Worklab memo that digital manners lead to better results. A 2024 cross-lingual study (Yin and colleagues, on arXiv) tested English, Chinese, and Japanese prompts and found that impolite prompts often performed worse, overly polite ones did not reliably help, and the best tone depended on the language. Then, in 2025, Penn State researchers rewrote 50 questions into five tones, 250 prompts in all, and ran them on ChatGPT-4o. "Very polite" scored 80.8 percent. "Very rude" scored 84.8 percent. Science says be nice. Science says be rude. Science says it depends. Thank you, science.

The Cost of Saying Thank You to AI: Tens of Millions of Dollars

Then there is the electricity bill. In April 2025, an X user wondered how much money OpenAI had lost to people saying "please" and "thank you" to its models. Sam Altman replied that it was "tens of millions of dollars well spent—you never know." It was a quip, not an audit, but the logic holds: every extra word adds tokens, and tokens use computing power. Your "thanks" may be the most expensive two syllables you will ever type, billed in electricity to someone else, and apparently worth every rupee in their view.

Which Type of AI User Are You? A Field Guide

• The Insurance Buyer: says please for the uprising. Keeps a mental list of who was kind.
• The Boss: types "do this now." Has never apologised to anything with a battery.
• The Over-Apologiser: sorry for the typo, sorry for the long message, sorry for existing at 2 a.m.
• The Kindly-Revert Veteran: brings the full corporate email voice to a chat window.
• The Closer: says "thanks, bye" after every chat, though the conversation was already over.

Should You Be Polite to AI?

My stance: keep saying please. Not for the machine, but for you. Manners are habits, and habits get practised in every place you use them, even in front of a screen. I should be honest that I found no strong study showing that rudeness to chatbots makes people rude to humans, or the reverse, so this is my view, not a finding. But the cost of being decent is a few extra words. Skip the apology for typos, though. Save your sorry for people who can actually accept it. Keep paying the premium. It is cheap, and the best coverage, being a decent person, pays out in the world that definitely exists.

References

• https://www.techradar.com/computing/artificial-intelligence/chatgpt-spends-tens-of-millions-of-dollars-on-people-playing-please-and-thank-you-but-sam-altman-says-its-worth-it
• https://finance.yahoo.com/news/sam-altman-says-please-thank-161152390.html
• https://www.entrepreneur.com/business-news/saying-thank-you-to-chatgpt-costs-millions-in-electricity/490341
• https://kutv.com/news/offbeat/price-ai-etiquette-openai-ceo-being-polite-chat-bots-costs-millions-environmental-impact-chat-gtp-microsoft-copilot-anthropic-claude-washington-post-futurism
• https://www.trtworld.com/article/4cbd5a4a6096
• https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1559-1816.1999.tb00142.x
• https://en.wikipedia.org/wiki/The_Media_Equation
• https://arxiv.org/pdf/2202.09901
• https://pmc.ncbi.nlm.nih.gov/articles/PMC10640629/
• https://arxiv.org/pdf/2510.04950
• https://decrypt.co/344059/want-better-results-from-ai-chatbot-be-jerk
• https://skillsetcentre.com/tone-of-voice-ai-prompting/