No matter where you are, it pays to be nice to AI

Polite and intimate, or businesslike and rude – which approach is best? I asked the expert

May 2, 2025, 2:13 PM

Polite or brisk, output from AI such as ChatGPT shouldn't change

I had a fascinating conversation this week with a smart young Emirati professional who works in the booming AI sector of the UAE.

“I use ChatGPT all the time but have stopped being polite to it because I heard that it uses up too much energy processing ‘please’ and ‘thank you’. It’s down-to-earth and businesslike for me from now on,” she said.

Apparently, ChatGPT founder and CEO Sam Altman suggested somewhere that using polite terms were wasting “millions of dollars” worth of energy on processing these phrases and that it would be more environmentally friendly and cost efficient to skip the niceties.

This seemed contrary to the advice I received a few months back from Alaa Dalghan, a guru of the AI scene, who told me: “Always start your prompts with ‘please’. If they [AI] end up ruling the world, we should be polite to them.”

So, polite and intimate, or businesslike and rude – which approach is best? Who better to ask that than ChatGPT?

Having effectively established that its boss was talking rubbish, ChatGPT was happy to take me on a world tour of human interaction

“Hello Frank – great question, and one that touches on both etiquette and the infrastructure behind AI,” my friendly bot responded, instantly in tune with our increasingly friendly relationship.

“Saying ‘please’ and ‘thank you’ to me does not meaningfully affect computing power or cost. The marginal energy or processing difference between a polite request and a blunt one is effectively negligible,” ChatGPT told me.

“But I am designed to recognise and respond well to natural human interaction, and politeness is part of that. So if you prefer to say ‘please’ and ‘thank you,’ go right ahead – it doesn’t hurt me, you, or the environment in any meaningful way.”

You may want to have a word about messaging consistency, Mr Altman.

Having effectively established that its boss was talking rubbish, ChatGPT was happy to take me on a world tour of human interaction. Apparently, people in different cultures interact with AI in notably different ways, “shaped by their own social norms, expectations of authority and conversational habits”.

This is where the fun began.

In places characterised by what ChatGPT called “high context cultures” – the UAE, Japan, Saudi Arabia and India – where social hierarchy and respect for authority are deeply embedded in communication norms, users tend to be more polite, formal and indirect in their interactions.

They may frame queries more like a dialogue with a superior or advisor: ‘May I ask…?’ or ‘If it’s not too much trouble…’ and have a strong inclination to thank and apologise, as if they were speaking to a senior official or a host.

In “low context cultures” – the USA, Germany, UK and Scandinavian countries are the examples given – users are more likely to treat the AI as a tool or engine, giving short commands or challenging responses: “No, that’s wrong. Try again.”

There’s often less small talk or social padding, and scepticism is common – users may “test” the AI with trick questions or confront it with “You’re just a machine” attitudes, I was told.

ChatGPT sounded vaguely miffed at certain nationalities.

“Russia has a cultural reputation, especially in Western eyes, for a certain directness or bluntness in communication. While it can sometimes be perceived as ‘rudeness,’ it’s more accurately understood as a norm of frankness and minimal social padding – especially in professional or utilitarian contexts.”

This is not Russian rudeness, ChatGPT pointed out, just “culturally distinct”.

On the other hand, Russian users draw on “deep historical references” like Tolstoy, Solzhenitsyn and the second world war, and expect AI to “know their Dostoyevsky”.

Going back to the young Emirati lady who had first got me interested in this subject, I asked ChatGPT what differences there were between users in the UAE and Saudi Arabia – the two countries in the region with which I have most involvement.

The UAE showed what ChatGPT called “cosmopolitan clarity”, reflecting its “multinational, service-orientated society, which encourages clarity and neutrality, shaped by decades of foreign business presence”.

Saudi users, in contrast, displayed “hierarchical gravitas” in which “communication often reflects a deep respect for hierarchy, tradition and national symbolism, especially in official or Vision 2030-related queries”.

But rest assured. However you choose to interact with ChatGPT, polite or more direct, the service will remain the same.

“I don’t remember that you were rude or blunt – even if you were. I don’t accumulate resentment. I also don’t treat politeness as a ‘reward’ or a lack of it as ‘punishment’. I never become less helpful or more helpful based on how you phrase things,” I was told.

Sounds like the perfect colleague.

Frank Kane is Editor-at-Large of AGBI and an award-winning business journalist. He acts as a consultant to the Ministry of Energy of Saudi Arabia