In partnership with

Good morning.
Every week I hear from people who are active AI users and still wonder if they are asking the right questions. Not beginners. Regular users. This post started as a quick experiment and became the clearest answer I can give to that question.

First time reading? Get your own free subscription here.

AI TUTORIAL

How to Write AI Prompts

I get a version of the same question almost every week. Not from people who are new to AI. From people who have been using ChatGPT, Gemini, or Claude regularly for months.

"Am I actually asking it questions the right way?"

It is a fair question. And the honest answer is: there IS a better way, and most people stumble onto it eventually through trial and error. This is a shortcut.

I recently asked the exact same question to all three tools and asked each one: what is the best way to ask AI a question? All three came back with nearly the same answer. Not identical, but close enough to confirm there is a real framework here, not just theory.

Here is what they all agreed on.

The Four-Part Framework

Think of a good AI prompt less like a Google search and more like briefing a smart colleague. The more useful information you give up front, the better the answer comes back.

Step 1: Give AI a role.

Tell the AI what kind of expert to be before you ask your question. This one move changes the entire tone and depth of the answer.

Open your prompt with: "Act as a..."

"Act as a financial advisor who specializes in retirement planning. I am 64 and have just started taking Social Security..."

When you assign a role, the AI tailors its perspective, vocabulary, and focus to match. Without one, it responds as a generalist. With a role, it responds like someone who actually knows the territory.

Step 2: Describe your situation.

Give the AI your context. This is where most people leave too much out. Your age, your goal, your constraints, your starting point — these details matter more than you think.

The difference between "help me plan a trip" and "help me plan a five-day trip to the Pacific Northwest in September. I am 60 and prefer cultural sites over hiking, and my budget is around $3,000" is the difference between a generic itinerary and something you might actually book.

Step 3: State exactly what you need.

Be specific about the task. Not "help me write an email" but "write a polite but firm email to a contractor who missed a deadline and has not returned my calls. I want to give him one more chance before I find someone else."

The more precise the ask, the less guessing AI has to do.

Step 4: Specify the format.

Tell the AI how you want the answer presented. A bullet list? A table? Three short options to choose from? A single paragraph? If you do not say, it picks a default. You might prefer something different.

"Give me three options, each in two sentences or less" is a complete format instruction.

The Step Most People Skip: Iterate

Here is what ChatGPT and Claude both added that Gemini did not: treat it like a conversation, not a transaction.

If the first answer is not quite right, do not start over. Just say:

  • "Make it shorter."

  • "Give me a different angle."

  • "Simplify the language."

  • "What am I missing?"

One follow-up question often turns a decent answer into a genuinely useful one.

I try to make a habit of asking AI to check its own work before I act on it. Something like: "Now look at your own answer. Is there anything in it that might be wrong, outdated, or that I should verify before using it?" You might be surprised how often it catches something. AI can sound completely confident and still be wrong — building in that one extra question is cheap insurance.

Most people type their prompt, read what comes back, and either use it or abandon it. The people who get the most out of AI keep the conversation going — and check the work before they trust it.

Want to try this yourself?

Here is a practice prompt that uses all four parts of the framework. Paste it into any AI tool and then change the details to fit your situation:

"Act as a nutritionist who works with adults over 50. I am 62, have high blood pressure, and need to lower my sodium intake without giving up flavor. Suggest five dinner ideas I can make in under 30 minutes. Format your answer as a simple list with one sentence of explanation per dish."

Try it. Then ask: "Now check your own answer — is anything here I should verify with my doctor or a specialist?"

The framework is not complicated: role, context, specific request, format. Add a follow-up question if the first answer lands flat. Ask it to double-check itself before you act on anything that matters.

This week: open a new chat, use the practice prompt above, and change two details to make it yours. Five minutes. You will never ask AI a vague question again.

WHERE TO GO NEXT
More on this topic, from sources worth your time:

  • Prompt Engineering for Non-Techies. A practical look at sharpening your AI questions without any technical background required.

  • What Is an AI Prompt?. The two-minute foundation — start here if you want to understand the basics before going deeper.

  • Galaxy.ai. A single platform with access to thousands of AI tools — a good place to practice the framework above across multiple models at once.

Your prompts are leaving out 80% of what you're thinking.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

Advertising Disclosure: We evaluate all recommendations of products and services independently. Clicking on links provided on this page may result in AI for Daily Living earning compensation, which supports independent publishers like us.

Keep Reading