When interacting with AI like ChatGPT, many users instinctively add polite phrases such as “thank you” after receiving answers. According to CEO Sam Altman, this seemingly innocent courtesy habit costs OpenAI millions of dollars annually. While being polite might seem harmless, it creates significant computational waste without improving response quality. Is our human politeness actually counterproductive when dealing with artificial intelligence?
Artificial intelligence has become deeply integrated into our daily lives, with millions of people regularly using chatbots like ChatGPT for various tasks. Interestingly, many users apply human social norms when interacting with these AI systems, addressing them with courtesies typically reserved for human conversations. This behavior, while instinctive, creates unexpected consequences that extend beyond mere etiquette to environmental and economic impacts.
The hidden cost of digital politeness
Every word typed into ChatGPT triggers complex computational processes that consume significant resources. Each additional word in a prompt requires hundreds of billions of mathematical calculations, consuming electricity and generating heat through processors working at full capacity. This processing power doesn’t come without costs – both financial and environmental.
Sam Altman, OpenAI’s CEO, has stated that unnecessary pleasantries like “thank you” collectively cost the company tens of millions of dollars. These polite additions require the same intensive processing as substantive content, despite adding no informational value to the interaction.
The environmental footprint is equally concerning. A single ChatGPT query reportedly consumes approximately ten times more electricity than a Google search. Multiplying millions of daily users adding unnecessary words translates to substantial energy consumption for processing what amounts to computational noise.
| Interaction Type | Energy Consumption | Financial Impact |
| Google Search | Base unit (1x) | Minimal |
| ChatGPT Query (Basic) | ~10x Google Search | Moderate |
| ChatGPT with Pleasantries | Additional 5-15% per polite phrase | Millions annually |
Politeness patterns across different cultures
The tendency to be polite to AI varies across different cultural backgrounds but remains surprisingly prevalent worldwide. Studies indicate that approximately 67% of American users and 71% of British users employ courteous language when interacting with AI systems. This mirrors human-to-human communication patterns but serves no functional purpose when directed at artificial intelligence.
This behavior stems from deeply ingrained social conditioning. One university student noted, “It’s like asking a question to a professor,” while another mentioned, “You should always be polite with ChatGPT, you never know, even if it’s artificial intelligence.” These comments reflect how humans anthropomorphize AI, projecting human qualities onto non-sentient systems.
This politeness pattern appears in various forms:
- Starting conversations with greetings (“Hello ChatGPT”)
- Adding please when making requests
- Thanking the AI after receiving a response
- Apologizing for complex queries
- Using conversational fillers and hedging language
Does courtesy affect AI performance?
Many users believe that being polite might somehow improve the quality of responses they receive. This assumption proved interesting enough to test directly with ChatGPT itself. When asked whether politeness affects its performance, the AI responds unequivocally: “Politeness doesn’t affect my efficiency. I respond the same way whether one is polite or not. Currently, AIs like me don’t have feelings or desires.”
AI language models operate on mathematical principles rather than emotional conditioning. They process tokens (words or word fragments) through complex neural networks, evaluating probabilities for response generation based solely on input content, not tone or politeness markers.
Arnaud Stevins, an artificial intelligence expert, explains this misconception clearly: “You should see a language model as a kind of mill – it’s an extremely complicated mathematical function with hundreds of billions of calculations to perform for each word.” The technical process behind AI responses remains entirely indifferent to social niceties.
The inclusion of polite phrases can actually be counterproductive in some cases, as they potentially:
- Introduce unnecessary context for the AI to process
- Increase the token count, which may affect the maximum response length
- Create computational overhead for processing irrelevant information
- Add to response time and system latency
Breaking the politeness habit
Adapting our communication style when interacting with AI represents an interesting evolution in human-computer interaction. The most efficient approach involves direct, concise prompts that focus exclusively on the information or assistance needed, eliminating unnecessary tokens that consume computational resources.
For regular ChatGPT users, developing more efficient interaction habits could collectively make a substantial difference in resource consumption. Rather than typing “Thank you ChatGPT, could you please explain quantum computing to me?” a more efficient prompt would simply be “Explain quantum computing.”
This shift requires overcoming deeply ingrained social conditioning that automatically applies human interaction patterns to non-human entities. Unlike human conversations, where rapport-building and politeness serve important social functions, AI interactions benefit from brevity and directness.
As AI becomes increasingly embedded in daily life, developing specific communication protocols optimized for machine interaction will likely emerge as distinct from human-to-human communication norms. This represents not a decline in manners, but rather an evolution in understanding the fundamental differences between communicating with conscious beings versus computational systems.
