⚠️ Affiliate Disclosure: This post is not an affiliate associated review. It may become one in the future and this disclosure will reflect that. This does not affect my reviews and I only recommend tools I genuinely believe in.
How Do You Write a Good AI Prompt?
You write a good AI prompt the way you would brief a capable new colleague who knows nothing about your situation: say what you want, why you want it, who it is for and what the finished result should look like. Most disappointing AI answers come from prompts that leave one of those out.
The tool barely matters. The same habits work in ChatGPT, Claude, Gemini, Copilot and the AI features built into apps like Canva and Grammarly. This guide covers the formula I use, real before-and-after examples, and what to do when the answer still is not right.
The Five Parts of a Good AI Prompt
You do not need all five every time. For anything that matters, though, run through the list before you hit enter.
- The task — what you want done, starting with a clear verb: write, summarise, compare, rewrite, plan, explain.
- The context — the background the AI cannot guess. Who you are, what it is for, what has already been tried, what the reader already knows.
- The audience — who will read the result. “For a customer who is annoyed” produces a very different email from “for my manager”.
- The format — length, structure and tone. “Three bullet points”, “a table with these columns”, “under 150 words, friendly but not chatty”.
- An example — if you have one, show it. A sample of the style or layout you want beats any description of it.
AI Prompt Examples: Before and After
| Weak prompt | Better prompt | Why it works |
|---|---|---|
| Write a LinkedIn post about our new product. | Write a LinkedIn post announcing our new invoicing app for UK sole traders. Audience: self-employed people who hate admin. Lead with the problem, not the product. Under 120 words, plain English, no hashtags in the body, one question at the end. | Adds audience, angle, length and tone |
| Summarise this report. | Summarise the attached report for our finance director, who has five minutes. Give the three findings that affect next year’s budget, each in one sentence, then one risk we should watch. | Defines the reader and what “summary” means to them |
| Give me ideas for a team away day. | Suggest five away-day activities for a team of twelve in Manchester, budget about £50 a head, mixed fitness levels, nothing involving alcohol. Put them in a table: activity, rough cost, why it works. | Adds constraints and a format you can act on |
None of the better prompts uses special wording or tricks. They simply answer the questions a sensible person would ask before starting the job.
How to Fix an AI Prompt That Gives Poor Results
Do not start again from scratch. Treat the first answer as a draft and give feedback, exactly as you would with a person:
- Say what is wrong with the AI’s answer, specifically — “too formal”, “the second point is wrong”, “cut it by half”.
- Ask it to ask you questions first — “before you answer, ask me anything you need to know” is one of the most useful prompts there is.
- Split a big request into steps — outline first, then each section, rather than a whole document in one go.
- Paste in the facts rather than relying on what the AI knows. It is the single best defence against AI hallucinations.
- Tell it what not to do — “no jargon”, “do not invent statistics”, “if you are unsure, say so”.
Common AI Prompt Mistakes
- Treating AI prompts like a Google search — three keywords give you a generic answer. Full sentences give you a useful one.
- Hiding the real goal — if the email is really meant to get a refund, say so.
- Asking for everything at once — ten requirements in one paragraph means some get dropped.
- Accepting the first answer — the second or third round is usually where the good version appears.
- Copying “magic prompts” — templates can be a starting point, but a prompt full of someone else’s context is worse than a short one with yours.
Do You Need to Learn Prompt Engineering?
For everyday use, no. “Prompt engineering” sounds technical, but for most people it simply means writing clear instructions and refining them. Modern AI models are far better at understanding plain requests than the early chatbots were, and many of the elaborate tricks that circulated in 2023 make little difference now.
What still makes a difference is the context only you can supply. That is the part no model can guess, and it is why two people using the same tool get such different results. My reviews of ChatGPT and Claude cover how each one responds to longer, detailed briefs.
Frequently Asked Questions
What is an example of a good AI prompt?
“Write a 100-word product description for a handmade oak chopping board, aimed at people buying a wedding gift. Warm, not flowery. Mention it is made in Yorkshire and needs oiling monthly.” It works because it gives the task, audience, tone, length and the facts the AI cannot know.
How long should an AI prompt be?
As long as it needs to be to give the context. A quick factual question can be one line; a piece of writing that matters might need a paragraph or more. Longer prompts are not a problem for modern AI tools — missing information is.
Do AI prompts work the same in ChatGPT, Claude and Gemini?
Mostly, yes. Clear tasks, context, audience and format help every major AI model. There are small differences in style and in how closely each follows formatting instructions, so it is worth refining a prompt the first time you use it in a new tool.
Should I say please and thank you to AI?
It does no harm and can make the tone of replies a little more natural, but it is not what makes a prompt work. Clear instructions and good context matter far more than politeness.
Can I save prompts I use often?
Yes. Most AI chat tools let you set custom instructions or project instructions that apply to every chat, so you do not have to repeat your role, preferred tone or formatting rules each time. Keep the reusable parts there and the task-specific detail in each prompt.
More AI Explained guides: What Is an LLM? · What Is a Token in AI? · What Is an AI Agent? · AI Hallucinations: Why AI Makes Things Up
James is an SEO and content specialist with over 12 years’ experience, and tests AI tools so everyday people don’t have to. WhatAITools.co.uk covers honest, no-jargon reviews to help you find the right tool for your needs.

Leave a Reply