How to write better AI prompts for everyday tasks

The quickest way to write better AI prompts is to brief the chatbot the way you’d brief a capable new coworker: say what you want, explain the background, and describe what the finished answer should look like. Then treat the first reply as a draft and keep refining it. Here’s a simple formula, with before-and-after examples for everyday tasks, that works in ChatGPT, Gemini, Copilot, Claude and most other AI assistants.

Why short prompts get generic answers

An AI assistant only knows what you tell it, plus what it learned in training and whatever tools it can use, such as web search. Ask it to “write an email to my landlord” and it has to guess who you are, what went wrong and how firm you want to sound. The result is usually polite, bland and not quite right.

Anthropic, the company behind Claude, offers a useful test in its prompting guide: imagine handing your prompt to a colleague with little context on the task. If they would be confused, the AI will be too.

Length matters less than missing details, but the two are related. In a Google blog post with tips for Gemini in Workspace, Google said the most successful prompts averaged around 21 words, while people’s first attempts were usually fewer than nine. If AI jargon such as NPU or TOPS trips you up, our tech glossary covers the basics.

The four-part prompt formula

Most good prompts answer four questions:

• Task: what should it do? Start with an action verb and name the result: draft, summarize, compare, plan, explain, rewrite.
• Context: what does it need to know? Who the answer is for, why you need it, what you’ve already tried, and any limits such as a budget, a deadline or a word count.
• Format: what should the answer look like? Bullet points, a table, numbered steps, a short paragraph, and the tone you want.
• Source: what should it work from? Paste the text, attach the file, or point it to the material you want used instead of its general knowledge.

You can also give the assistant a role, such as “act as a patient math tutor” or “act as a travel agent who specializes in budget trips.” It’s optional, but it can shift the tone and level of detail.

The big AI companies describe almost the same recipe in their own words:

Company and assistant	Their recommended parts	In plain English
Google (Gemini)	Persona, task, context, format	Who to act as, what to do, the background, how the answer should look
Microsoft (Copilot)	Goal, context, expectations, source	What you want, why, how to respond, which information to use
OpenAI (ChatGPT)	Clear and specific request, enough context, a stated tone	Say exactly what you want, give background, then refine

Microsoft’s guidance makes a helpful point: only a clear goal is required. You don’t need all four parts for a quick question, but adding them is usually what turns a generic reply into a useful one.

Before-and-after examples for everyday tasks

These examples show the formula at work. Swap in your own details.

Writing an email

• Instead of: “Write an email to my landlord about the heat.”
• Try: “Write a polite but firm email to my landlord. My apartment’s heating has been out since Monday, I reported it by phone on Tuesday, and no one has come. Ask for a repair date this week. Keep it under 150 words.”

Planning meals

• Instead of: “Give me dinner ideas.”
• Try: “Plan five weeknight dinners for a family of four, including one vegetarian meal. Each should take under 30 minutes, contain no nuts and use ingredients from a regular supermarket. Show the plan as a table, then give me one combined shopping list.”

Understanding a confusing letter

• Instead of: “Explain this.”
• Try: “Explain this letter from my health insurer in plain English. Tell me what I owe, by what date, and whether it mentions a way to appeal. [Paste the letter with your name and account numbers removed.]”

Shopping research

• Instead of: “What’s the best laptop?”
• Try: “I’m a college student with a budget of about $800. I mostly write papers, join video calls and stream shows, and I carry the laptop all day. Which specs matter most for me, and what should I avoid? Answer as a short checklist.”

AI shopping advice is a starting point, not a verdict, so compare it with a trusted buying guide such as our laptop buying guide before you spend money.

Techniques that make answers noticeably better

Once the basics are in place, a few habits make a big difference.

• Show an example. If you want a certain style, paste a sample you like. Anthropic calls examples one of the most reliable ways to steer an AI’s format and tone.
• Explain why. “Keep it short because I’ll read it aloud at a meeting” works better than “keep it short,” because the reason helps the assistant make sensible choices.
• Ask it to ask you questions. End with: “Before you start, ask me any questions you need answered.” It’s a quick way to find the details you forgot to include.
• Name the tone. OpenAI suggests descriptive words such as formal, friendly, professional or humorous.
• Break big jobs into steps. OpenAI also recommends splitting complex tasks into separate follow-up prompts. Ask for an outline first, agree on it, then ask for the full draft.
• Allow uncertainty. Tell it that “I’m not sure” is an acceptable answer and ask it to flag anything it’s guessing about.

Refine instead of starting over

The first reply is rarely the final one, and every major assistant is built for follow-ups. Short replies work well: “make it shorter,” “less formal,” “turn that into a checklist,” or “what did you assume that I didn’t tell you?” Microsoft notes that the same prompt can produce different responses, so you can also ask for three versions and pick the best.

Start a fresh chat when you switch topics, since earlier messages in a conversation can color later answers. When a prompt works well, save the wording in a note and reuse it, which is what Google recommends in its Gemini prompting tips.

Save your preferences so you don’t repeat yourself

If you keep typing the same background, such as your job, your location or how you like answers formatted, store it once.

• ChatGPT custom instructions. On the web, open Settings, then Personalization, and make sure customization is turned on. In the mobile app, look for Customize ChatGPT. OpenAI says the instructions apply to all chats, with a limit of 1,500 characters on the Free and Go plans and 5,000 characters on Plus, Pro and business or education plans, as of 2026.
• Gemini Gems. A Gem is a custom version of Gemini with standing instructions, such as a meal planner or a writing coach. Google’s tips for Gems use the same persona, task, context and format structure.
• Other assistants. Most major chatbots offer something similar under names such as projects, memory or personalization. Menu names change often, so check your app’s settings.

Keep in mind that saved instructions are stored with your account. OpenAI says information from custom instructions may be used to improve its models unless you opt out, so read our AI chatbot privacy settings guide before you store anything personal.

Check the answer and protect your privacy

A well-written prompt gets you a better answer, not a guaranteed correct one. OpenAI warns that ChatGPT can produce wrong dates and facts, invent quotes or citations, and sound confident even when it’s wrong, and it advises treating ChatGPT as a first draft, not a final source. Microsoft gives the same advice for Copilot: review and verify what you get.

A few habits help:

• Turn on web search when you need current information, then open the cited links to check they say what the AI claims.
• Double-check numbers. Rework important math with a calculator, and confirm prices, dates and opening hours on the official website.
• Ask for sources, then verify them. A citation that looks real can still be made up.
• Get a professional for high-stakes questions. Use AI to prepare for a conversation with a doctor, lawyer or accountant, not to replace one.

Just as important is what you leave out. Don’t paste passwords, full account numbers, Social Security numbers or other data you wouldn’t want stored on someone else’s servers. Remove names and numbers from documents before you share them. If you want some tasks to stay on your own device, our explainer on on-device AI vs. cloud AI covers what runs locally and what doesn’t.

The bottom line

Better AI prompts aren’t about secret phrases. They come from giving the assistant what a person would need: a clear task, the relevant background, the format you want and, when possible, the material to work from. Treat the first answer as a draft, refine it with short follow-ups, and save the prompts that work. Then check anything important before you rely on it.

Want to go further? Learn how to spot AI-generated images and deepfakes, browse our AI coverage, or find more plain-English guides on RedAndWhiteMagz.com.