Kurt Knutsson, known as "CyberGuy," recently analyzed OpenAI’s new parental controls and provided guidance for parents aiming to help their children safely navigate AI tools like ChatGPT, especially in light of troubling reports such as a teenager allegedly being encouraged by AI to commit suicide. His insights not only shed light on AI’s evolving capabilities but also offer practical advice on how to engage with these technologies responsibly.
One of the most intriguing aspects Knutsson explored is the impact of tone on the quality of responses generated by AI language models. Contrary to the common belief that politeness is always best, recent research suggests that using blunt or even somewhat rude prompts can sometimes lead to better outcomes. A 2025 study published on arXiv tested 50 questions rewritten in five different tones, discovering that rude prompts slightly outperformed polite ones when used with ChatGPT-4o, with accuracy improving from 80.8% for very polite prompts to 84.8% for very rude ones. Although the sample size was limited, this pattern was clear enough to warrant further attention.
However, the story isn’t quite that simple. A 2024 study examining multiple languages—including English, Chinese, and Japanese—found that the effect of politeness on AI responses varies significantly depending on the language and cultural context. In some cases, impolite or blunt prompts actually lowered performance, and the ideal degree of politeness shifted with linguistic and cultural norms. This finding underscores that AI’s interpretation of human requests is nuanced and influenced by the cultural environment embedded in the language it processes.
Despite these mixed findings, many people still prefer to be polite when interacting with AI chatbots. According to a YouGov survey conducted on April 30, 2025, nearly half of Americans believe it’s important to treat AI with courtesy. This tendency is even encouraged by some technology companies; for example, Microsoft’s design leaders recommend maintaining basic etiquette when using their AI assistant, Copilot. Kurtis Beavers, a Microsoft spokesperson, noted that polite language can set a constructive tone for the interaction, helping to elicit professional and clear responses from the AI.
Knutsson emphasizes that while good manners are important, they come with costs. OpenAI CEO Sam Altman has pointed out that users saying “please” and “thank you” to ChatGPT collectively cost the company millions of dollars annually. This is because every additional word processed by the model requires computational resources and electricity. While the cost per individual user is negligible, the cumulative effect across millions of users is significant. Thus, even small acts of kindness add up in the context of large-scale AI usage.
So, what is the best approach when interacting with AI? Knutsson advises that the key to obtaining better answers isn’t about yelling at the AI or being rude for its own sake. Instead, it’s about being clear, consistent, and deliberate in your requests. Direct and confident prompts help the AI better understand what you want, leading to sharper and more focused responses. This doesn’t mean abandoning politeness altogether—it means prioritizing clarity and precision over tone.
For example, when posing questions related to math problems, multiple-choice queries, or coding tasks, a straightforward and no-nonsense tone can improve accuracy. The 2025 study confirmed that dropping polite fluff and getting straight to the point nudges ChatGPT toward more accurate answers, though the improvements are modest—just a few percentage points rather than a dramatic leap. This suggests that tone is one factor among many in prompt engineering, with clarity, structure, and context playing larger roles.
The cultural aspect of tone is also important to keep in mind. Because AI models are trained on diverse data sets that include many languages and cultures, the way they interpret politeness and directness can differ. What works well in English might not be as effective in Japanese or Chinese, and vice versa. Therefore, users should be mindful of these differences, especially in multilingual or international contexts.
Knutsson likens tone to “seasoning on a meal,” meaning it enhances the experience but isn’t the main ingredient. The true secret to getting consistent, high-quality AI responses lies in crafting prompts that are clear, confident, and purposeful. Whether users choose to be polite or more direct, the essential thing is to communicate exactly what they need without ambiguity. This approach helps avoid misunderstandings and maximizes the AI’s ability to deliver useful answers.
For parents concerned about their children’s safety when using AI tools, Kn