Have you ever asked an AI something and it gave you the wrong answer? Many times you tend to give an answer that is too long for something simple, associates it with other things or has hallucinations.
For example, you could ask Copilot to help you to free up space on your PC and give you more details than necessary or complicate things. The same can happen with generating images in ChatGPT, because maybe you want your photo to become Ghibli style and it gives you another one that has nothing to do with it.
But is it artificial intelligence’s fault? In itself, everything revolves around the prompts and their structure. Simple questions are not enough to give you everything you need and that is why it is crucial to know how to do them.
Luckily, there are general rules that work for Claude, Gemini, DeepSeek, Grok, and most chatbots. If you think you have problems getting what you need from these virtual assistants, then it is time for you to know the following tricks.
Give all the necessary and specific context
The most common mistake that many users make is that they simply talk to the artificial intelligence as if it were another person or as if it were a teacher being asked direct questions.
It is not enough to ask you to resolve a question. or perform work of any kind. For each of the things, you have to take the time to add the necessary context and be specific depending on the situation.
For example, asking «do an essay on the jobs that are going to be replaced by AI» is not the same as saying «I need a report with a professional tone that covers what artificial intelligence is, why it is a risk for human jobs and what can be done to avoid becoming unemployed. Add an introduction and conclusion. Let each section have three paragraphs and be a total of 2,000 words.»
As you can see in this prompt, you say what you require, the tone, the sections and the number of titles, paragraphs and words it should have. This is of utmost importance to better adapt to what you needespecially for text generation or creativity in writing.
As for the images, it is similar, although I will share with you a trick that always works for me and that is the formula Object/Person + Context/Description + Setting + Style + Format. Use this on most adding the details and you will see it completely change.
Let’s say you’re looking for an image of a laptop. It wouldn’t be «make a laptop», but rather «Create an image of a laptop with Windows 11 on a wooden desk. Let the background be a brick wall and scattered pencils on the desk. Let it look hyper-realistic and have a 16:9 aspect ratio (1280×720).»
Determine constraints for the response
When you make the prompt, it is good that you consider indicate restrictions to prevent him from saying something specificdeviates from information or adds inappropriate things.
To that same formula you could mention that “don’t remember this” so that it avoids storing the response in ChatGPT or Gemini memory. Also, you could ask him to remove certain fillersspecific symbols or other types of fragments that you don’t need.
Another important thing you should know is that after it gives you that answer, most wizards allow it to regenerate with a circled arrow icon. Additionally, it is possible to edit that information or add other things that may have been missing.
Use the role-playing tactic
The «role playing» strategy is one of the most useful and works for practically everything you need. It’s basically ask the AI to play the role of a professional by starting the prompt with «You are an expert on» or a specific character like Steve Jobs, Elon Musk, and Bill Gates.
Thus, the chatbot will act and respond according to the capabilities offered by that expert, so will give you valuable information. If you need a routine for the gym, I would be a coach and so on with everything else.
Put each of these tips into practice and you will see that using this new technology is not so complicated after all. It is certain that you will obtain better results when applying them in your daily life.