If you've ever asked an AI assistant a question and gotten back a confident but utterly useless answer, the problem usually isn't the model — it's the prompt. Prompt engineering is the skill of asking in a way that gets you what you actually want. It's quickly becoming a core literacy, and the good news is the fundamentals are simple enough for anyone to learn. This is the first in a short series; here we cover what prompt engineering is and the single biggest lever you control: specificity.
So what is prompt engineering?
A prompt is just the instruction you give an AI model. Prompt engineering is the practice of crafting and refining that instruction to reliably get a useful result. It isn't coding, and it isn't magic words. At its heart it's clear communication — with one twist: the thing you're communicating with can't see your screen, doesn't know your goal, and won't ask follow-up questions unless you invite it to.
A few ideas do most of the work:
- Clarity — say exactly what you want, with no room for guessing.
- Context — give the background the model needs to be relevant to your situation.
- Constraints — set the guardrails: length, format, tone, audience.
- Iteration — treat the first answer as a draft and refine, rather than expecting perfection on attempt one.
Why "the model can't read your mind"
An AI model predicts a likely response based on the words you give it and the patterns it learned during training. If your words are broad, the most likely response is also broad — generic, safe, and not particularly helpful. You didn't get a bad answer; you got the average answer to a vague question. Specificity is what pulls the model away from the average and toward your actual need.
Here's the same underlying question asked three ways. Watch how the quality of the answer tracks the quality of the prompt:
- No context
- Too broad
- Not actionable
- Some context
- Narrows the answer
- Still lacks depth
- Clear goal & audience
- Asks for a format
- Actionable answer
Same question, three rungs of specificity. The answer is only ever as good as the rung you stand on.
The four ingredients you added without noticing
The jump from the first prompt to the third wasn't luck. Four ingredients did the heavy lifting, and you can add them deliberately every time:
- Who you are — "I'm a beginner with no coding skills" tells the model what level to pitch at.
- What you want — "publish a one-page portfolio site" is a concrete outcome, not a vague topic.
- The constraints — "for free" rules out whole categories of answer that would waste your time.
- The format — "step-by-step instructions... what to do at each step" shapes the answer into something usable.
It's a conversation, not a vending machine
The biggest mindset shift is to stop treating the first answer as final. A vending machine gives you one shot; a good assistant improves when you react to it. "That's too technical — explain it for a 10-year-old." "Shorter, just the three key points." "Now give me an example." Each correction is itself a prompt, and refining across a few turns almost always beats agonising over the perfect opening line.
This is why "prompt engineering" has the word engineering in it: not because it's difficult, but because it's iterative. You observe the output, adjust the input, and converge on what you need — the same loop an engineer uses to tune anything.
The takeaway
Prompt engineering isn't a secret handshake. It's clarity, context, constraints, and a willingness to iterate. The fastest way to get better results from any AI tool — today, with no new software — is to climb one rung up the specificity ladder before you hit enter. In the next article we'll turn this into a practical checklist of habits you can apply to every prompt you write.