Showing the AI what you want with examples

In this guide, you will learn how showing the AI an example of what you want can get you a far better result than describing it in words.

Telling versus showing

There are two ways to ask an AI for something. You can describe what you want, or you can show it. Both work, but they suit different jobs, and knowing when to reach for each is a quiet superpower.

Describing is what most people do by default. You write a prompt like “write me a friendly product description” and let the AI decide what “friendly” means. Because you have given it no example to copy, this is sometimes called zero-shot prompting: zero shots, because you are giving zero examples to work from.

Showing is the other way. Instead of describing the style you want, you paste in a sample of it and say “more like this.” Giving the AI one or two examples to copy is called few-shot prompting, and for anything with a particular voice or shape, it is often the shortest path to a result you are happy with.

Why an example beats a description

Style is genuinely hard to put into words. If someone asked you to describe your own writing voice, you might reach for “warm” or “clear,” but those words mean different things to different people, and they certainly mean something different to an AI.

An example skips the guessing. When you show the AI a paragraph you like, it can pick up on things you would struggle to name: the rhythm of the sentences, how formal it is, whether it uses questions, how it opens and closes. It matches the feel, not a rough description of the feel.

This is why “here is one I wrote earlier, do another like it” works so reliably. You are handing the AI a target to aim at instead of a blindfold and good wishes.

How many examples do you need?

Usually one is enough to shift the result noticeably. Two or three help when the pattern is subtle or when you want the AI to spot a rule you have not spelled out, like “notice these all start with a question.”

You do not need a perfect example, either. A rough sample of the right shape beats a polished description of the wrong one. If all you have is something close, share it and tell the AI which parts to keep and which to change.

Try it yourself

Pick a small writing task where voice matters, like a social media caption or a short bio. First ask for it plainly, then ask again with an example attached. Here is the second version:

Write a short caption for a photo of our new bakery loaf.

Match the style of these two captions we've used before:
"Fresh out of the oven and gone by lunch. Sourdough, Tuesdays only."
"We proofed this one overnight so you don't have to. Grab it while it's warm."

Keep it to one or two sentences, playful, no hashtags.

Send both versions and read them side by side. The example-led one will sound far more like you, because you gave the AI something real to copy rather than a word like “playful” to interpret.

A common first instinct

A common first instinct is to describe the style you want in ever more adjectives, piling on “warm but professional, friendly but not too casual” and hoping the AI lands it. You will get there much faster if you stop describing and start showing: paste one example of the real thing and say “like this.” A single good sample does more work than a paragraph of instructions.

Next steps

You can now aim the AI with examples, which is one of the most reliable moves in prompt craft. Next, we will look at a handful of higher-leverage techniques in advanced prompt techniques, where examples become one tool among several.