The debate over whether to be polite to AI chatbots has taken an unexpected financial turn. Last week, OpenAI CEO Sam Altman revealed that the simple acts of saying “please” and “thank you” to ChatGPT cost the company tens of millions of dollars in electricity. Yet, he described this expense as “well spent.”
This revelation reignites a debate I explored last year for TVNewsCheck in The Surprising Reasons Why You Should Be Nice To AI Chatbots. Research has shown that being nice to a chatbot, like ChatGPT, will generate better results for the user. But the debate has a new twist. Now, the question isn’t just about should you be nice to your chatbot; it’s about whether politeness is worth its price tag, even for local media companies. As media companies continue to deploy AI tools across their organization, they are now faced with a new unpredictable budget line: AI tokens. And every word or phrase, including “thank you” and “please” uses these tokens.
The Case For Politeness
Better Outputs, Fewer Biases: Studies from institutions including Waseda University and the RIKEN Center found that GPT-3.5 performed 15.6% better with courteous inputs, while GPT-4 showed reduced bias when users framed requests thoughtfully. So, statistically speaking, if you are mean to a chatbot, it won’t work as hard for you. But polite language aligns with the AI’s training data, and will help prioritize coherent, helpful replies.
“At scale, small words add up,” says Ben Gerst, CTO at NOTA. “Politeness in this context may seem trivial, but it shapes how AI systems respond and evolve. If we’re building systems that learn from every prompt to be assistive, then the way we speak to them isn’t just etiquette – it’s training data.”
A Cultural Shift In AI Interaction: A Future PLC survey of 1,000 users revealed that 70% are polite to AI, with 12% cutely admitting they do so “in case of a robot uprising.” But this wave of politeness isn’t just quirky human behavior, it is significantly influencing the ways in which AI systems are programmed to respond and evolve with human users. By prioritizing politeness and positive interaction, users are not only creating a mutually beneficial experience, but also potentially guiding the ethical frameworks within which AI operates.
The Cost Of Kindness
Energy and Environmental Impact: Every extra word processed by an AI model uses AI tokens and consumes energy and generates compute costs . Multiply that by millions of sessions, and the financial and environmental toll becomes undeniable. Critics argue that pouring resources into courtesy for emotionless machines is wasteful. “We’re in the middle of deploying our AI enterprise platform, called AskMorgan, to hundreds of users across the company,” says Colin Benedict, VP of news at Morgan Murphy Media. “At scale, even minor adjustments to usage behavior and the models attached to our apps can add up to significant longterm cost savings.”
A Pragmatic Middle Ground
Some experts advocate for “utility politeness” — just enough courtesy to improve outputs without excess.
“With AI, clarity trumps courtesy,” suggests Trajan Bayly, founder and CEO of Gray Falkon. “While politeness has its place in human interaction, when it comes to AI, precision and directness are what truly drive results. The key is to communicate your intent clearly and concisely, without unnecessary embellishments.”
Prioritizing Adoption
As many media companies are in the process of deploying their AI enterprise solutions, is burdening users with restrictions around politeness even worth worrying about?
“Right now, even something as simple as adding polite or contextual phrasing can influence an AI model’s response to better match the tone you’re aiming for,” says Michael Newman, director of transformation at Graham Media Group. “We are not steering people away from that—we actively encourage it. In our training, we emphasize the importance of giving models more context because the cost of an extra sentence is far lower than the cost of getting a useless answer.”
Shaping Future AI
Today’s interactions train tomorrow’s models. Polite exchanges could lead to AI that defaults to helpfulness, while curt ones might reinforce transactional behavior.
“The minimal cost of politeness is a worthwhile investment in both technological efficiency and human behavioral patterns,” says Erica Neiges, global supply chain head at Amplo Global. “The behaviors we practice repeatedly tend to become habits, regardless of who or what we’re interacting with. How we communicate with AI reinforces our broader communication patterns and reflects the collaborative mindset we need to nurture to address complex global challenges.
A Question Of Values
Altman’s stance suggests OpenAI views politeness as a feature, not a bug. But as a society we must weigh:
- Performance vs. cost: Does a better reply justify the energy expense?
- Habit formation: Will polite AI interactions improve human communication — or erode it?
- The software itself: Maybe we need to rethink how we build AI software?
“When politeness to a chatbot carries such a high price tag, maybe we should be rethinking the design — not the users.” says Helen Todd, co-founder and CEO, Creativity Squared.
Being nice to your chatbot is an interesting debate. But with hard numbers from Altman himself, we must confront a balancing act: optimizing performance, minimizing utility waste and deciding what kind of digital society we’re building for the future.
“Tens of millions well spent” might sound ridiculous, but it underscores a deeper truth: How we talk to AI shapes how it talks back. And if the AI models we create today become the learning blocks for the AI of our future, isn’t it all worth it?
Disclosure: Ordo Digital is consulting with Graham Media Group and Morgan Murphy Media on AI enterprise deployment projects.
