Prompt a Better Prompt

There was a time when searching Google meant typing a few keywords into a box and hoping the right page appeared. As search improved, many of us became comfortable asking a full question, or even describing what we needed in a paragraph.

AI calls for another shift. It can respond to a short request, but it can do much more when you explain what you are trying to accomplish. The phrase “prompt engineering” may make that sound like a specialist’s job. It is not. Lee Boonstra’s Google guide to prompt engineering covers advanced techniques, but its most useful starting lessons are simple: be clear, give relevant context, show examples, and refine what you ask. You do not need to memorize a secret formula. You need to give the AI enough direction to be useful.

Think of a prompt as a brief you would give a capable assistant. “Write a marketing email” gives that assistant very little to work with. Who is the email for? What are you offering? What should the reader do next? What tone sounds like you? A little thought before you hit Enter can save a great deal of revision afterward.

Start with the job

Begin with a clear action: summarize this proposal, compare these two options, draft a reply, or suggest five names for a new service. Then say why you need the result.

Suppose you run a local roofing company. You might start with: “Write a social media post about our roof inspections.” That will produce something. A more useful prompt might be:

“Write a short Facebook post for our local roofing company offering free roof inspections before winter. Our customers are homeowners in the area, many with older roofs. Keep it friendly and straightforward. End by inviting them to call to book an inspection this month.”

The second version gives the AI a job, an audience, facts to work with, a tone, and a goal. It is longer because the details matter, not because longer prompts are automatically better.

Show what good looks like

If you have an example of a post, email, or description that sounds right, include it. One example can tell the AI more about your taste than a string of adjectives. You can say, “Use this as a guide to the tone, but write something new.”

You can also specify the shape of the answer. Do you want three options or one polished draft? A paragraph or a checklist? Plain language or a technical explanation? If you know how you will use the result, say so.

You do not have to supply every detail at once. But the details you already know should not be left for the AI to guess.

Say what you do not want

Good direction includes boundaries. In Google Ads, advertisers choose keywords that may trigger an ad, and negative keywords that keep it from appearing in the wrong searches. A prompt can work in a similar way: tell the AI what belongs in the response and what should be left out.

For the roofing post, you might add: “Do not promise same-day service or quote prices. Avoid exclamation marks and scare tactics about leaks.” That is more useful than hoping the AI will somehow know your standards.

Keep the positive instruction in the lead. Tell it what success looks like, then add the exclusions that matter. A long list of prohibitions without a clear goal can leave the AI unsure what to produce.

Let the AI improve your prompt

You do not need to write a perfect prompt from scratch. Start with the rough version you have and ask the AI to help strengthen it:

“Here is my basic request. Rewrite it as a clearer prompt. Show me what details are missing, and do not invent facts about my business.”

There are dedicated prompt-enhancing tools and websites that do this too. You can begin with the AI you already use. Read the suggested prompt before you send it. A more elaborate prompt is only better if it still reflects your real goal and does not smuggle in assumptions you never intended.

Ask it to ask you

One of the simplest additions to a prompt is a question at the end: “Before you begin, what do you need to clarify? Is there anything important I have forgotten to tell you?”

That turns a one-way request into a chance to catch gaps. Perhaps the AI needs to know the price, the audience, a deadline, or whether the message is for existing customers or new ones. Answering those questions may improve the result more than any clever wording could.

When the first response arrives, read it critically. Correct the facts, point to what works, and ask for a revision where needed. Prompting is an experiment: try a version, examine the result, and adjust.

From prompting to conversation

Once you have made a thoughtful request, you do not have to treat the first answer as the end. Read it as the beginning of a working exchange. You can say, “That is close, but the tone is too formal,” or “Where did that claim come from?” You can ask the AI to compare two approaches, explain an assumption, or try again with a detail you forgot to mention.

This is how my own use of AI began to change. In 2024, I experimented with playing Twenty Questions. I was not just looking for a game to pass the time. I wanted to see whether the AI could follow the rules, keep track of what I had said, and adapt as the exchange went on. It gave me a simple setting in which to practice a different skill: having a dialogue with AI instead of expecting one perfect answer to one perfect prompt.

You do not need to start with Twenty Questions. Pick a small, low-stakes task and stay curious about the response. Ask why it chose a particular suggestion. Tell it what it misunderstood. Give it another piece of context and see whether its answer improves. If it asks you a useful question in return, answer it. Each turn can make the next one more useful.

That does not mean trusting every reply. I still test what AI tells me. I question claims, look for errors, and notice when it takes too long, loses the thread, or sounds confident without good reason. If a fact matters to a customer or a business decision, I check it. Experimenting with AI includes finding its limits as well as its strengths.

The techniques in this article work whether you type or speak. Start by giving your next request a better brief: the job, the context, what good looks like, and what to avoid. Invite questions about anything you missed. Then, when the answer arrives, keep going.

Prompting helps you get answers. Conversation can help you discover questions you had not thought to ask.

You do not need to be a prompt engineer to prompt a better prompt.