How to Write Better AI Prompts
Turn good prompts into a repeatable system.
Use one repeatable method to turn a rough request into a clearer, reusable instruction. Then test the result, identify what failed, and adapt the prompt for the task and model.
This guide turns the earlier basics and examples into a method you can reuse. Continue with mistakes and advanced techniques before the LLM deep-dive guides.
The 6-element better prompt formula
Each relevant element can reduce a specific ambiguity, but unnecessary or conflicting instructions can make a prompt worse. Use only what the task and evaluation criteria require.
Basic → Better → Advanced
The same topic at three levels of instruction detail. Each version adds one or more formula elements; the effect varies by task, model, and source quality.
| Basic prompt | Better prompt | Advanced prompt |
|---|---|---|
| Write about coffee | Write a 700-word blog post about the health benefits of coffee for busy professionals. Include 5 benefits with H2 headings. | Act as a nutrition content editor. Write a search-optimised article on coffee benefits for office workers aged 25–45. Include H2 headings, caution notes, a FAQ, and a CTA. Under 900 words. No promotional language. |
| Write social media post | Write 3 LinkedIn posts for a freelance designer announcing a new logo service. Professional but approachable tone. | Act as a B2B copywriter. Write 5 LinkedIn posts for a freelance designer launching a new brand identity service targeting early-stage startups. Use the hook-insight-CTA format. Each post under 150 words. Avoid generic phrases like "excited to announce". |
| Email customer | Write a friendly support email explaining a 2-day refund delay. Under 100 words. End with an apology and a solution. | Act as a senior customer experience manager. Write an empathetic support email for a customer whose refund is delayed by 2 days due to a bank processing issue. Acknowledge frustration, explain simply, offer a concrete next step, and include a goodwill gesture. Under 120 words. No corporate filler. |
| Explain AI | Explain artificial intelligence to a complete beginner in 5 bullet points. Use simple language and one real-world analogy. | Act as a science communicator. Explain what artificial intelligence is to a 14-year-old with no technical background. Use 3 analogies from everyday life. Avoid the word "algorithm". Format: short intro, 3 analogy blocks, one sentence conclusion. Under 200 words. |
| Make it shorter | Rewrite this paragraph to be 40% shorter. Keep the main argument. Remove filler phrases and passive voice. | Act as a senior editor. Rewrite the paragraph below to be exactly 50 words. Rules: keep the central argument, remove all passive voice, cut every sentence that doesn't move the argument forward, keep the original tone. Show the original word count and the new word count. |
Apply focused improvements, build a prompt live, and save the structures that work for repeated tasks.
7 techniques to upgrade any prompt
Each technique is a single focused change you can apply to any prompt you already have. Pick the one that fits what you're trying to improve.
Words like "write", "help", "improve", and "explain" are starting points, not instructions. Replace them with exact deliverables — what type, how many, how long, in what format.
A role can steer perspective, vocabulary, priorities, and tone. It does not add verified qualifications or knowledge, so use it only when that lens helps and check factual claims independently.
Models can be more verbose than the task needs. A relevant word or sentence limit can encourage prioritisation, but overly tight or conflicting limits may omit important content or reduce quality.
Exclusion rules can prevent predictable style problems such as generic openings, jargon, filler phrases, or hollow enthusiasm. Keep the list short and make sure exclusions do not conflict with required content.
A relevant example can communicate a pattern that is difficult to describe. Use representative examples, remove sensitive data, and check that the model has not copied unwanted wording or errors — or adapt one from these AI prompt examples.
When a draft has useful parts, use precise follow-up instructions to say what to keep and what to change. Restart or change sources when the draft is built on a false premise or unreliable evidence.
For analytical or strategic tasks, provide explicit criteria and ask for assumptions, trade-offs, a concise rationale, and a verification checklist. Some reasoning models prefer direct goals over "think step by step", so follow model-specific guidance.
Live prompt builder
Fill in any combination of fields and watch a structured prompt assemble in real time. Copy it directly into any AI.
Build your prompt library
A useful AI productivity habit is to save prompts that perform well on representative tasks. Record the model, date, inputs, and quality criteria because prompts can behave differently after model or product changes.
Example library categories
Use this meta-prompt to turn a useful prompt into a reusable starting template, then test it on representative inputs.
FAQ
How do I know if a prompt is good?
How many formula elements should I use?
What is a prompt template?
How do I build a personal prompt library?
Does the AI tool matter for prompt quality?
Related Prompting Guides
Strong prompts become easier to reuse when you keep a library of structures that performed well in your own tests. Revisit AI prompt examples and adaptable templates whenever you want templates you can adapt and evaluate.
Build your prompt library next
Take everything from this series and put it into practice. Next, create a structured prompt with a generator, then apply it in a workflow such as marketing, website building, studying, or business communication.