AI Doesn't Care About Your 'Please' And 'Thank You'

Forget Your 'Pleases' And 'Thank Yous' AI Just Doesn't Care

The notion that polite requests or flattery can somehow optimise AI chatbot interactions has become a pervasive myth. From imagining AIs as discerning Starship captains to showering them with praise, much popular advice on ‘prompt engineering’ crumbles under rigorous examination.

It's time to debunk these theatrical fantasies and focus on what truly makes a difference.

Recent research, for instance, investigated whether “positive thinking” genuinely enhanced AI chatbot accuracy. Experimenters labelled AIs “smart,” urged careful thought, and even concluded prompts with “This will be fun!”

Yet, none of these tactics consistently improved performance and often wasted valuable computational resources.

Intriguingly, one specific technique did produce a surprising, albeit whimsical, result: making an AI pretend it was commanding a Starship Enterprise crew actually boosted its basic mathematical prowess. While certainly an anomaly, this highlights the often unpredictable and non-human logic governing AI responses.

“A lot of people think there's some magic set of words you can use that will make LLMs solve a problem,” says Jules White, a computer science professor at Vanderbilt University. “But it's not about word choice, it's about how you fundamentally express what you're trying to do.”

The Costly Charade of AI Etiquette

In 2025, a user on X (formerly Twitter) posed a poignant question: “I wonder how much money OpenAI has lost in electricity costs from people saying 'please' and 'thank you' to their models.” Sam Altman, OpenAI's CEO, offered a somewhat cryptic retort: “Tens of millions of dollars well spent. You never know.”

While the “tens of millions” figure is likely anecdotal, it certainly underscores a widespread perception. Many interpreted Altman’s “You never know” as a cheeky nod to future AI existential risks, yet the question of politeness also carries very practical implications for resource consumption and training data, which ultimately translates into real-world costs for providers like OpenAI.

Large Language Models (LLMs) function by dissecting your input into “tokens,” which are then statistically analysed to forge responses. This implies that every minute detail, from word selection to punctuation, influences the AI’s output.

The real challenge, however, lies in the sheer unpredictability and non-obvious nature of these influences; politeness is often just noise.

Conflicting studies are rife concerning the impact of minor linguistic variations. For instance, a 2024 study – while quickly contested – suggested that LLMs offered superior, more accurate answers when prompted politely rather than with blunt commands.

Curiously, cultural differences surfaced; Japanese-speaking chatbots actually performed marginally worse when users were overly courteous compared to their Chinese and English equivalents. Many observers in the Asia-Pacific region are keenly watching how AI models adapt to diverse linguistic and cultural nuances, considering the region's linguistic diversity.

“Politeness may not protect you from angry robots or make LLMs more accurate, but there are other reasons to keep doing it.”

However, don't rush to draft a thank-you note for your AI just yet. Another informal test indicated a previous version of ChatGPT was more accurate when subjected to insults. Crucially, the research landscape is relatively nascent, and AI models are evolving at such a breakneck pace that findings can become obsolete astoundingly quickly.

This continuous evolution means established techniques can swiftly lose their efficacy, a point often overlooked when discussing AI reliability in rapidly developing Southeast Asian markets like Singapore or South Korea. This constant state of flux makes AI a dynamic, yet challenging, field to master, particularly regarding prompt effectiveness, as detailed in AI's Blunders: Why Your Brain Still Matters More.

Smarter AI Communication: Practical Strategies

Experts now largely concur that newer, more sophisticated AI models from entities such as OpenAI, Google, and Anthropic are considerably less susceptible to superficial cues like flattery or insults. Techniques that once appeared to yield better results, including the whimsical Star Trek role-play, are frequently rendered redundant by enhanced model sophistication. For more on the inner workings of such companies, consider "I’m deeply uncomfortable with these decisions" - Anthropic's CEO.

Key takeaway: AI tools are mimics, not sentient beings. They simulate human behaviour without possessing genuine emotions or understanding. To elicit superior responses, abandon the anthropomorphic approach and treat the AI as the sophisticated tool it is.

Understanding this distinction is vital for optimising your interactions and obtaining more useful outputs.

Here’s a breakdown of current best practices for effective AI prompting, moving beyond superficial pleasantries:

• Ask for Multiple Options: Rather than a singular answer, instruct the AI to generate three to five varied options. This encourages critical evaluation and refines your comprehension. For example, when summarising a report, request several perspectives or lengths.
• Provide Examples: If you desire a specific writing style from the AI, don't just list instructions. Furnish genuine samples of your preferred output. “Here are 10 emails I've dispatched; please replicate this tone and structure,” is far more potent than generic stylistic directives.
• Initiate an Interview: For complex tasks, prompt the AI to query you iteratively. When creating a job description, instruct the AI: “Ask me questions, one by one, until you have sufficient information to draft a compelling listing.” This adaptive method yields more tailored and accurate results.

The Enduring Human Element

Interestingly, human politeness towards AI endures, despite its practical irrelevance. A 2025 survey by a publisher reported that 70% of individuals are polite to AI, largely because they deem it “the right thing to do.”

A smaller, yet significant, 12% confessed to doing so out of a perceived hedge against potential robot uprisings – a notion far more common in Western pop culture than in serious policy discussions across APAC regions, where regulations frequently focus on data privacy and ethical development rather than sentient AI threats.

Rick Battle and other experts confirm that modern AI models, particularly those featuring in leading products like ChatGPT, Gemini, or Claude, are far better at discerning the core intent of your prompt. They are less likely to be swayed by superficial linguistic flourishes in any consistent or exploitable manner.

So, while politeness might not render your chatbot “smarter,” it could enhance your comfort and ease of interaction, making the process more enjoyable, much like understanding 5 Ways Google Gemini Is Changing How Students Learn can make education more accessible.

As the philosopher Immanuel Kant argued regarding cruelty to animals, being unkind to anything can be damaging to one's own character. While an AI cannot have its feelings hurt, maintaining courtesy can be a beneficial personal habit. So, the real question isn't whether AI needs your manners, but what personal boundaries or standards do YOU uphold when interacting with advanced AI systems?

Drop your take in the comments!