We’ve all been taught from a young age that manners don’t cost a thing. Say please. Say thank you. Don’t interrupt. Don’t feed the AI after midnight (wait, wrong horror story).

But recently, that old saying is getting a bit of a rethink, especially when it comes to AI. Yep, our pleasantries might be costing the planet.

Let’s rewind a sec. Over the past few weeks, there’s been a bit of a stir about how the language we use with large language models (LLMs) like ChatGPT can have very real energy implications. Sam Altman, OpenAI’s CEO, reportedly said that people using manners - saying "please" and "thank you" in their prompts - add up in cost and compute, and those polite little extras are part of why running these tools is, well… eye-wateringly expensive.

So, should we drop the niceties?

Look, we’re not saying you need to start barking orders at your friendly neighbourhood chatbot like you’re trying to assert dominance over a rebellious smart fridge. But it does raise a few interesting questions.

Every word you type into an AI prompt (every syllable, emoji, and yes, even every “pretty please”) has to be processed, encoded, and interpreted. And all that computation? It happens in energy-hungry data centres.

Training GPT-3 (just training, mind you, not running it daily) is estimated to have used about 1,287 MWh of electricity and generated over 550 metric tons of carbon of electricity and generated over 550 metric tons of carbon dioxide. That’s the same as flying a plane from New York to San Francisco 550 times. ?

And that’s just one model. Add in the day-to-day usage from millions of users worldwide, and the impact becomes a little harder to shrug off.

AI vs. the Planet: Who’s Winning?

There’s a growing concern about AI’s carbon footprint. According to the International Energy Agency, data centres already account for around 1–1.5% of global electricity use, and that number is climbing... fast.

As AI tools become more accessible, more powerful and more integrated into daily workflows (hello, AI-enabled everything), we’re collectively pulling more juice from the grid than ever.

Do Manners Actually Help?

Some people argue that politeness can actually improve your results with AI. It’s like social engineering for robots; asking nicely seems to nudge the models toward more detailed, thoughtful answers. No, it’s not magic, but think of it like soft prompt engineering. Politeness = better tone = sometimes more useful output.

Some early testing from users shows that prompts with positive or cooperative language tend to reduce the chances of abrupt, unhelpful, or "I can’t do that" style replies. It’s not a perfect science, but there’s something there.

TL;DR?

• Yes, every word you type into AI systems gets processed and adds to the overall energy load.
• Training a big model like GPT-3 can emit more CO₂ than five average cars over their entire lifetime .
• Being overly verbose or unnecessarily polite (yep, we’re going there) might add to that cost.
• But! Manners might help improve the tone and helpfulness of AI responses.
• And if/when the machines rise up, better to be on their good side, right?

So what’s the verdict?

Our take? Don’t lose sleep over the odd "please" or "thank you." But if you’re sending a novella of niceties every time you want to know how to update your firewall rules, maybe rein it in a tad. Keep it sharp, clear, and efficient.

Still, there’s no harm in a little charm. After all, if we’re going to be coexisting with AI, it doesn’t hurt to start the relationship on good terms.

Let us know in the comments: Are you on Team Manners or Team Max Efficiency?