Introduction
Artificial intelligence (AI) like ChatGPT is a powerful tool, but it's still important to know how to use it properly. Many users complain about inaccurate or irrelevant responses. The key? Knowing how to write your prompts wellIn this article, we will give you all the keys to mastering the art of prompting and fully exploiting the potential of generative AI, while reducing its environmental impact.
Why is AI polluting? ?
Artificial intelligence requires powerful infrastructure to function. Each request sent to a model like ChatGPT mobilizes energy-consuming servers located in data centers that consume large amounts of electricity and water for cooling. According to some studies, the carbon footprint of an AI can be equivalent to hundreds of Google searches for a single complex interaction.
Misuse of AI therefore leads to:
- Overconsumption of energy due to the multiple requests required to obtain a satisfactory response.
- An increased carbon footprint, particularly because of data centers powered by fossil fuels.
- A waste of IT resources, avoidable by optimizing queries.
By writing precise and well-formulated prompts, you reduce the number of necessary exchanges and optimize the use of resources, thus contributing to more responsible AI.
Writing a good prompt is essential!
AI generates content based on the instructions it receives. If you give it a vague instruction, the response will be just as vague. A good prompt allows you to:
- getting precise and relevant answers.
- Win time avoiding multiple iterations.
- Improve quality and creativity some answers.
- Reducing the energy consumption of AI avoiding unnecessary queries.
The 6 golden rules for a good prompt
1. Be specific and detailed
A quick wave like “Give me some UX advice” is too broad. Specify your request:
→ “Give me 5 UX tips to improve the accessibility of an e-commerce site.”
2. Indicate the expected format
AI can generate responses in different forms: list, structured text, table, etc.
→”Give me a comparison table of UX best practices for a mobile vs desktop site.”
3. Provide context
AI has no memory between sessions. Bring context to your query.
→ “Imagine you’re a UX design expert specializing in SaaS. Explain to me the most common mistakes in user onboarding.”
4. Use examples
Explaining what you expect by providing examples helps refine the answer.
→ “Generate a catchy slogan for an eco-friendly clothing brand. Example: 'Dress Consciously.'”
5. Detail what you don't want to see
AI doesn't know what's useless or unsuitable for you unless you tell it.
→ “Propose a UX audit of a mobile banking application, but without generic recommendations or overly obvious UX principles.”
6. Experiment and iterate
If the answer is not satisfactory, reword your prompt by adding more details or changing the approach.
→ “Show me the same answer but in bullet point form and more concisely.”
Tips:
Think of AI as a colleague. It can't know what's in your head and it has no memory. If we take the example of ChatGPT, AI has no memory in the sense that it would remember previous exchanges once the conversation is over. At the start of each conversation, ChatGPT starts “from scratch”.
This means that the model does not keep any information about past exchanges previous conversations to refine their responses. In other words, each interaction is treated independently, with no history or context other than that provided at the present moment.
However, during an ongoing conversation session, ChatGPT is able to track and remember information shared within that same discussion. This allows it to tailor its responses, respond more appropriately, and better follow the thread of the conversation.
That's why, for your idea to be understood, you must be as specific as possible.
Advanced tips for ultra-efficient prompts
Play with roles → “Imagine you are a designer at Apple, how would you improve this interface?”
Constrain the AI → “Give me an answer in less than 100 words.”
Ask for multiple perspectives → “Suggest 3 different UX solutions to improve the navigation menu.”
Conclusion
We often tend to use ChatGPT like a Google search engine, out of habit. But these requests have an impact, not only on the environment but also on the time you spend refining your requests. The more time you take to build your prompt from the start, the more your search will be tailored to your needs and the more time you will save.
More sources:
ChatGPT: How to Write a Good Prompt in 2025?
As Use of AI Soars, So Does the Energy and Water It Requires
Getting started with prompts for text-based Generative AI tools | Harvard University Information Technology
How to Write AI Prompts: The Key to Better Outputs from Generative AI – Contently
Anaëlle Staelen, UX/UI designer and Product designer at UX-Republic
