How to Write Better AI Prompts: A Beginner's Guide

Most people who try AI and find it underwhelming have one thing in common: they are writing vague prompts and getting vague results. The quality of your output depends almost entirely on the quality of your input.

This guide will teach you how to write prompts that get genuinely useful results from ChatGPT, Claude, Gemini, or any other AI assistant.

What Is a Prompt?

A prompt is any instruction, question, or input you give to an AI tool. It can be a single sentence or several paragraphs. It can include examples, documents, or context. The AI reads your prompt and generates a response based on it.

If the response is unhelpful, the solution is almost always to improve the prompt, not to assume the AI is incapable.

Why Prompt Quality Matters So Much

AI language models are incredibly capable, but they are also incredibly literal. They respond to what you write, not what you meant to write. Poor prompts produce outputs that are generic, off-tone, too long, too short, or simply miss the point.

Good prompting is the most transferable skill in AI. It works across every tool and every task. Investing 30 minutes in understanding it will pay dividends every single day.

The RACE Framework for Better Prompts

RACE stands for: Role, Action, Context, Example. Use it as a mental checklist when writing any significant prompt.

R: Role

Tell the AI who it should be for this task. "You are a senior marketing copywriter" or "You are an experienced Australian HR manager." This activates the right domain knowledge and tone.

A: Action

Be specific about what you want. Not just "write an email" but "write a 150-word follow-up email." Not just "summarise this" but "summarise this in 5 bullet points, focusing on action items."

C: Context

Provide relevant background. Who is the audience? What is the purpose? What are the constraints? The more relevant context you give, the more targeted the output.

E: Example

Show the AI what a good result looks like. Paste in an example email, document, or piece of writing that matches your desired style. This is the fastest way to nail tone and format.

5 Before-and-After Examples

Example 1: Writing a client email

Before: "Write an email to a client who has not paid their invoice."

After: "You are a professional services consultant in Melbourne. Write a polite but firm 100-word email to a client whose invoice for $4,500 is 14 days overdue. Keep the tone professional and relationship-preserving. End with a clear call to action."

Difference: The second prompt produces something you can actually send. The first produces a generic template.

Example 2: Summarising a document

Before: "Summarise this report."

After: "Summarise this 20-page market research report into 5 key findings for a non-technical executive audience. Each finding should be one sentence with a supporting data point. Focus on findings relevant to the Australian retail sector."

Difference: The second prompt produces something boardroom-ready. The first produces a generic overview.

Example 3: Social media caption

Before: "Write a LinkedIn post about AI training."

After: "Write a 100-word LinkedIn post from the perspective of a Melbourne business owner who recently completed an AI training workshop with their team. Tone: authentic, practical, slightly surprised at the results. Include one specific result (hours saved per week). End with a question to drive engagement."

Difference: The second prompt produces a post that sounds human and specific. The first produces generic corporate speak.

Example 4: Brainstorming

Before: "Give me ideas for marketing my business."

After: "You are a growth marketing consultant. I run a 10-person accounting firm in Brisbane serving small businesses with turnovers of $500K to $5M. Give me 8 specific marketing ideas to generate new client enquiries in the next 90 days. Prioritise low-cost tactics. Avoid vague suggestions like 'improve your social media': be specific."

Difference: The second prompt produces 8 actionable tactics you can start tomorrow. The first produces a generic list you could find anywhere.

Example 5: Writing a job ad

Before: "Write a job ad for an office manager."

After: "Write a 300-word job ad for an Office Manager at a 15-person digital marketing agency in Sydney. The role involves managing calendars, suppliers, and office operations. Salary: $70,000 to $80,000. We value people who are proactive and calm under pressure. Tone: friendly and direct, not corporate. Avoid cliches like 'fast-paced environment.'"

Difference: The second prompt produces something you can post immediately. The first produces a bland template that will attract the wrong candidates.

Common Mistakes to Avoid

• Being too vague: "Help me with marketing" tells the AI almost nothing. Be specific about task, audience, format, and length.
• Forgetting the audience: Who will read this? The AI defaults to a generic audience unless you specify.
• Accepting the first draft: The first output is usually good but rarely perfect. Ask the AI to revise: "Make it shorter," "Change the tone to be more direct," "Add one more example."
• Not providing examples: If you have a style you love, paste it in. AI is excellent at matching style when shown what you want.
• Over-complicating it: Some tasks need one sentence. Not every prompt needs all four RACE elements. Start simple, then add detail if the outputs miss the mark.

How to Iterate

Think of prompting as a conversation, not a one-shot command. If the first response is not right, refine it:

• "That is too formal. Rewrite it in a more conversational tone."
• "Cut this by half and make the key point clearer."
• "You missed the point about [X]. Revise with that as the central focus."

You can also ask the AI to improve your prompt: "I will give you a task. First suggest how I should prompt you to get the best result, then complete the task."

For structured prompting workshops, see the courses at Top AI Academy.

Read next: ChatGPT vs Gemini vs Claude: Which AI Should You Use?

---

Frequently Asked Questions

Do I need to learn coding to write good prompts?

No. Prompt engineering for everyday business tasks is plain English. You are writing instructions, not code. Technical prompting (for API use and system prompts) requires more knowledge, but for general business use, clear writing is the skill that matters most.

Is there a limit to how long my prompt can be?

Each AI tool has a context window, measured in tokens (roughly 3/4 of a word each). ChatGPT-4o handles around 128,000 tokens. Claude 3.5 Sonnet handles 200,000 tokens. For practical business prompts, you are unlikely to hit these limits. You can include long documents for summarisation or analysis.

Can the AI remember what I told it last time?

Within a conversation, yes. Across different sessions, no, by default. ChatGPT and Claude both offer memory features (paid plans) that allow some context to persist between sessions. For business use, consider using consistent system prompts or project instructions to re-establish context.

What is the difference between a prompt and a system prompt?

A system prompt is a set of instructions given to the AI before the conversation starts, usually by a developer or power user. It sets the AI's persona, constraints, and knowledge base. A regular prompt is what you type in the chat. For most business users, you only need to worry about the regular prompt.