When someone says that an AI “is useless” or “gives generic answers,” in most cases the problem is not the AI. It’s how they asked the question.
Modern language models are extraordinarily capable tools, but they are not mind readers. They respond to the text they receive. And if that text is vague, ambiguous, or incomplete, the response will be too. The good news is that learning to write better instructions requires no technical knowledge. It requires clarity of thought.
Why most prompts fail
The most common prompt looks something like: “Write me an email about the project.” Or “give me ideas for my business.” Or “explain how the stock market works.”
These instructions fail for the same reason: they are too broad to produce anything useful. The AI doesn’t know which project, who the email is for, what tone you want, how long it should be, or what you’re trying to achieve. It also doesn’t know what kind of business you have, what stage it’s at, what you’ve already tried, or what resources you have. Nor does it know how much you already know about stocks, which aspect you want to understand, or why you need it.
Without that information, the model does the only thing it can: fill in the gaps with generic assumptions. And the result is, indeed, generic.
Another frequent mistake is asking for several things at once in a single prompt. “Analyze this, give me three conclusions, write it in a formal but approachable tone, and keep it under a paragraph.” Too many simultaneous criteria usually produce text that meets all of them halfway.
The four elements of a good prompt
An effective prompt doesn’t need to be long. It needs to be precise. There are four elements that, when present, dramatically improve the quality of the response.
Context. Who are you, what are you trying to do, what is your situation? The more relevant information you provide, the more calibrated the response will be. “I’m the communications director of a B2B technology company” produces different responses than “I’m a freelancer selling design services to small businesses.”
Specific task. Not “give me ideas,” but “give me five content ideas for LinkedIn targeting HR managers at companies with 50 to 200 employees.” Specificity reduces the margin for interpretation and increases usefulness.
Desired format. If you want a list, say so. If you want running prose, say so. If you want the AI to ask you questions before responding, say so. AI doesn’t read minds, but it follows format instructions with great fidelity.
Constraints. Length, tone, level of detail, what NOT to include. Constraints are as important as positive instructions. “Don’t use technical jargon” or “in fewer than 150 words” narrow the response space toward what you actually need.
Real examples: bad and good prompts
Bad: “Write me a social media post about productivity.”
Good: “Write a 150-word LinkedIn post about managing interruptions while working from home. Direct and practical tone, no motivational clichés. The audience is people working remotely with young children. Start with a concrete situation, not a rhetorical question.”
The difference is not that the second is longer. It’s that the second leaves no room for unnecessary assumptions.
Bad: “Explain artificial intelligence to me.”
Good: “Explain how large language models (LLMs) work as if I were someone with a biology background but no computer science knowledge. Use analogies with biological processes where possible. Maximum three paragraphs.”
The first prompt could generate anything from a Wikipedia entry to a technical tutorial. The second defines the level, the focus, and the length.
Context is the most underrated ingredient
Of the four elements, context is the one most frequently omitted and the one with the most impact.
Language AIs have no access to your particular situation unless you give it to them. They don’t know if you’ve been at your company for two years or two months. They don’t know if your client is skeptical or enthusiastic. They don’t know if you’ve already tried something and it didn’t work.
The more relevant context you share, the more the AI can act as an informed collaborator rather than a generic text generator. You don’t need to write endless paragraphs. It’s enough to include the details that would change the response if the model knew them.
A useful technique: before writing your prompt, ask yourself “what would a very intelligent person with no knowledge of my situation need to know to give me a genuinely useful answer?” That is exactly what should be in the prompt.
A good prompt is not one that gives perfect instructions. It’s one that eliminates enough ambiguity for the response to be actionable.
How to improve with practice
The skill of writing good prompts develops like any other: by doing it and paying attention to results.
When a response isn’t what you expected, instead of rewriting from scratch, ask yourself which element was missing. Was it context? Was the task not specific enough? Was the format undefined?
You can also use the AI itself to improve your prompts. Tell it: “This is my prompt. What additional information would you need to respond more usefully?” Or: “What ambiguities do you see in this instruction?” Good models are quite capable of pointing out the gaps.
With time, writing a good prompt takes the same 30 seconds as a bad one. The difference is in the result. And in the time you won’t waste revising and rephrasing responses that don’t work.
