Being polite to AI might make you feel better—but it may also be costing OpenAI more money. A new analysis suggests that longer prompts filled with pleasantries and conversational padding may be subtly increasing the computational load—and therefore the operational expenses—of running large language models like ChatGPT.
The idea, originally highlighted in a blog post by machine learning engineer Lior Sinclair, has gained traction online. Sinclair calculated that small talk and polite phrasing may be adding up to thousands—or even millions—of extra tokens processed daily, especially as OpenAI scales its services for enterprise users.
“When millions of users say ‘please’ and ‘thank you’ to ChatGPT, that translates to millions of unnecessary tokens,” Sinclair wrote. “Each token costs real compute.”
🧠 Politeness in Prompts = More Tokens
Large language models like GPT-4 (and now OpenAI’s O3 model) process inputs by counting “tokens”, which are chunks of words. The more tokens a prompt includes, the more compute power and cost is required.
With OpenAI charging enterprise clients based on token usage, even small increases—such as adding “Hello, can you please…” instead of just issuing a direct command—can inflate usage and costs over time.
“Multiply a few extra tokens by millions of daily prompts, and you’ve got a budget line item,” a former OpenAI engineer told TechCrunch.
💬 Is It Really a Problem?
While the cost-per-token is small, it adds up fast at scale. Some analysts believe OpenAI could be spending millions per year on what is effectively digital small talk. This is especially relevant in high-volume deployments like:
- Customer support chatbots
- Coding assistants like Copilot or ChatGPT Team
- API integrations for SaaS platforms
Even small optimization in how prompts are written—internally or by end users—could reduce costs or speed up response times.
Still, many users defend the habit of treating AI politely, either for the sake of user experience or to encourage more ethical relationships with technology.
“Teaching politeness, even to a machine, reinforces good social behavior,” one educator told TechCrunch.
🤖 What’s OpenAI Saying?
OpenAI has not issued a formal statement on the matter, though developer documentation does note that token efficiency can improve performance and pricing. Internal teams are reportedly looking at ways to auto-trim “soft prompts”before they are sent to the model engine—without affecting perceived friendliness.
