Generative AI: The Art of the Prompt

We’ve already talked about what generative AI is and how it can carry on a conversation that feels surprisingly natural — much like speaking with a real person. But something that’s absolutely crucial is this: to get good answers, you need to ask good questions — or more precisely, write good prompts.

Over time, I’ve been experimenting and improving the way I request content (and I say “content” because it’s not just limited to chat — it also applies to image generation). So I’d like to share 10 important tips to help you write better prompts:

1 - Start by asking yourself

Before overwhelming the AI with a confusing question, try asking it to yourself first and see if you can answer it. This not only helps clarify the question, but might also help you refine it and make it more specific if your initial answer feels too broad.

2 - Keep it simply simple

Sometimes what we want to ask is inherently complex, which can make our prompt tangled or unclear. But the great thing about AI is that you can always ask follow-up questions. So there’s no need to start with something long or overly complicated — begin with a simple prompt and build from there.

3 - Iterate

Tied to the previous point: understand that getting the right answer is often a process, not something that happens on the first try. If the answer isn’t quite right, go back to step one: ask yourself, simplify the question, and try again.

4 - Context is key

Just like when we talk to people, providing extra context helps get better answers. For example, it’s not the same to ask for information to present to a formal committee as it is to gather notes for personal use. This even applies to the tone, formality, or region-specific phrasing in the response.

5 - Set boundaries

Using the example above — if you ask for a summary of a 1,000-page book, it could be way too long… or too short. It’s important to set upper and/or lower limits to help the AI hit the mark. For instance: “Give me a summary between 100 and 200 words.”

6 - Ask for examples

One of the most common uses of chat-based AI is explaining topics we don’t understand. And the best way to understand something is with real-world examples — so make sure to ask for not just an explanation, but also use cases, sample scenarios, or practical applications.

7 - Format the response

If you know who the response is intended for, specify the best format. You can even ask it to respond as if it were written by a celebrity or a historical figure. The content might be the same, but the style of presentation can make a big difference.

8 - Use negation to your advantage

This is especially useful in image generation. For example, you can tell the AI not to include text in an image, or not to show any people. This helps the model narrow down your request and produce more accurate results — faster and with fewer attempts.

9 - Ask for sources

This becomes especially important when you’re using the AI to form conclusions or gather facts. If you’re going to reference the output, you should understand where the information came from. So don’t hesitate to ask it to cite its sources.

10 - Show me the numbers

In addition to references, it’s often helpful to back up responses with metrics or statistics. For example, if you ask which programming languages are most used, the answer should include percentages — since the difference between first and second place might be tiny or huge, and that could change your entire perception.

Have you tried asking a generative AI anything yet? What other tips would you add to the list?