Table of Contents
Most people use a large language model (an LLM, the AI behind ChatGPT, Claude and Gemini) like a search box. They type a short question, get a decent answer, and move on.
That works for trivia. It doesn’t work for the jobs that eat your week, like client emails, vendor comparisons, project plans and hard conversations.
The difference is almost never the model. It’s the prompt. A thin request gets a generic answer, while a prompt with a reader, a limit and a format gets something you can use.
Below are ten prompts for LLMs that I’d keep in a saved note. Each one is filled in with a real scenario, so you can see how much detail a good prompt carries, and each works in ChatGPT, Claude or Gemini. Swap in your own details and they’re ready.
What Makes a Good Prompt for an LLM
Every prompt below follows the same simple logic, and it’s the same logic the big AI companies teach.
Anthropic’s guidance says Claude responds well to clear, explicit instructions. Google’s guide for Gemini in Workspace breaks a good prompt into persona, task, context and format. Microsoft uses goal, context, expectations and source for Copilot.
The wording differs, but the idea is the same. Say what you want, say who it’s for, give the background, and describe the shape of the answer. The ten prompts below do exactly that.
10 Prompts for LLMs You Can Copy and Use Today
1. The Brief-First Prompt
Use it for any writing task where tone matters. The trick is to give the model a role, a reader and a clear limit before it writes a word.
You are a customer success manager at a small web design agency. Write an email to a client named Daniel explaining that his website launch will slip from 12 November to 26 November because the photographer delivered images late. Daniel is anxious about the deadline and has a marketing campaign planned for December.
Keep it under 150 words. Own the delay without blaming the photographer. State the new date clearly in the first two sentences. End with one concrete step Daniel can take to help us stay on schedule. Tone: calm, warm and direct.
2. The Ask-Me-Questions-First Prompt
When you don’t know what to include, let the LLM interview you. This beats guessing at a long prompt and often surfaces things you hadn’t considered.
I want to plan a one-day workshop for 20 small-business owners on using AI for marketing. Before you suggest anything, ask me up to seven questions, one at a time, that you need answered to plan it well. Wait for my answer to each question before asking the next.
Focus on what would change the plan: the audience's experience level, budget, venue, time available, and what attendees should leave with. When you have enough information, say so and write the plan.
3. The Explain-It-at-My-Level Prompt
LLMs explain things in a default middle register. Tell them who you are and they adjust.
Explain how email deliverability works to a bakery owner who sends a monthly newsletter and has never heard the term. Use one everyday analogy. Avoid jargon, or define each term in plain words the first time it appears. Keep it under 250 words.
Then give me three quick questions to check that I understood, and wait for my answers before telling me which ones I got right.
4. The Skeptical Editor Prompt
Most people ask an LLM to “improve” their writing and get a rewrite that loses their voice. This prompt asks for critique instead.
You are a skeptical editor. Read the draft pasted below, a 300-word "About Us" page for a freelance bookkeeping business. Tell me the single biggest weakness. Then quote three specific sentences that are vague or generic, exactly as written, and explain why each one falls flat. Finish with one thing that works well.
Do not rewrite anything. I'll revise it myself.
Paste your draft right after it. Because the model has to quote your own sentences, it can’t hide behind vague feedback.
5. The Summarize-for-a-Decision Prompt
A summary is only useful if it serves a decision. Tell the model what you’re deciding and what to look for.
The document pasted below is a 12-page vendor proposal for new accounting software. I need to decide by Friday whether to take it to my business partner. Summarize it in under 200 words, focusing on total cost (including any fees that aren't obvious), contract length, what happens if we cancel, and any claim that sounds like marketing instead of a commitment.
End with three questions I should ask the vendor. Use only what's in the document, and if something important is missing, say so.
6. The Messy-Notes-to-Table Prompt
LLMs are very good at turning scattered notes into structure. The key is telling them what to do with gaps.
Below are my notes from calling four suppliers about printed packaging. Turn them into a table with these columns: supplier, price per 1,000 units, minimum order, turnaround time, and notes. If a detail wasn't mentioned, write "not stated" instead of guessing. Keep prices in the currency I gave.
After the table, tell me which supplier looks cheapest and which looks fastest, based only on the table.
The “not stated” rule matters more than it looks. Without it, models tend to fill empty cells with plausible numbers.
7. The Compare-Options Prompt
Decisions get clearer when the criteria are named up front. This prompt also makes the model show its assumptions, so you know what to verify.
I'm choosing between two project management tools for a five-person design studio: Trello and Notion. We mainly need client approval tracking and simple task boards. We have a small budget, and nobody wants to spend time on setup.
Compare the two on setup effort, price for five users, client-facing features and long-term flexibility. State any assumption you're making about current pricing or features, because I'll verify them. End with a recommendation in one short paragraph, and name the one situation in which you'd choose the other option.
8. The Break-It-Down Prompt
Big goals stall because nobody has turned them into small steps. Give the model your real time limits and it will build a plan you can actually follow.
I'm launching an online course in eight weeks while working full-time. I can spend about six hours a week, on weekday evenings and Sunday mornings. Break the launch into tasks that each take 30 to 90 minutes, group them by week, and order them so nothing depends on a task that comes later.
Give a time estimate for each task. Flag the three tasks most likely to slip. Leave the final week lighter as a buffer before launch day.
Splitting work like this is also official advice. Google’s guide recommends you break related tasks into separate prompts instead of cramming everything into one.
9. The Show-Your-Uncertainty Prompt
LLMs can sound equally confident whether they’re right or wrong. This prompt forces them to mark what needs checking.
Write a 200-word overview of the main differences between a sole proprietorship and an LLC for a freelancer in the United States. After the overview, list every factual claim you are less than fully confident about. For each one, tell me what kind of source I should check, such as a government website or an accountant.
Don't hide uncertainty to sound authoritative.
Why bother? One study asked ChatGPT to write literature reviews and found that 55% of GPT-3.5’s citations and 18% of GPT-4’s were entirely fabricated. Newer models have improved, but the habit of checking still pays off, especially for legal, medical or financial details.
10. The Practice-the-Conversation Prompt
This is the most underused one. Before a tough conversation, rehearse it with the model playing the other side.
Act as a client named Priya, who runs a boutique gym and has just told me my freelance rate is too high. I'm a freelance social media manager charging 600 dollars a month. Stay in character: be polite but push back realistically, use two or three common objections, and don't give in easily.
After each of my replies, respond as Priya. After five exchanges, step out of character and give me feedback: what I handled well, where I sounded defensive, and one stronger phrase I could have used.
How to Adapt These Prompts to Your Own Work
You don’t need all ten. Pick the two or three that match your week and rewrite them with your own facts.
Three habits make the biggest difference. First, always name the reader, because it changes the output more than almost anything else. Second, add a rule for gaps, such as “if something isn’t stated, say so.” Third, set a limit on length so you don’t get three paragraphs when you need three sentences.
Avoid the common slips too. Don’t paste a prompt without changing the details, don’t send confidential client data to a tool your workplace hasn’t approved, and don’t publish anything without checking the facts.
Specific Prompts Beat Clever Ones
You don’t need tricks to get good answers from an LLM. You need a clear task, a real reader, sensible limits and a format you can use.
Try one prompt from this list today, on a task that’s already on your plate. Compare the result with what you’d have gotten from a one-line request, and you’ll see why the extra thirty seconds are worth it.