How to Write Better AI Prompts — 8 Techniques That Work Every Time

Most people type a question into an AI tool the same way they'd type it into Google. The results are mediocre — and they assume AI isn't that useful. The problem isn't the AI; it's the prompt. These 8 techniques work on ChatGPT, Claude, Gemini, and any other text AI.

Why Prompts Matter So Much

AI language models are instruction-following machines. They generate the most probable continuation of whatever you write. A vague prompt gets a generic, averaged response. A specific prompt with clear context gets something genuinely useful.

The same AI, the same model, the same question — different results depending entirely on how you phrase it.

1. Give It a Role

Start with "You are a [role]" to frame the perspective and tone of the response.

Weak: "Explain investing to me"

Better: "You are a financial educator who specialises in explaining investing to complete beginners. Explain index funds to someone who has never invested before, using simple analogies. Avoid jargon."

The role tells the AI the expertise level to draw from and the tone to use.

2. Specify the Format

Tell the AI exactly what you want the output to look like. Without this, you'll get a generic paragraph response when you might want a table, bullet points, or a step-by-step list.

Format options you can request:

• "Give me a numbered list of 7 items"
• "Format this as a table with columns: Tool, Cost, Best For"
• "Write this as a step-by-step guide with a heading for each step"
• "Give me bullet points — no more than one sentence each"
• "Write this as a professional email, subject line included"

3. Set a Word or Length Limit

AI tools default to medium-length responses. Being specific about length gets you what you actually need.

• "In 3 sentences or fewer" — for quick answers
• "In exactly 150 words" — for social media captions, bios, or constrained formats
• "Give me a comprehensive 800-word guide" — when you actually want depth

4. Provide an Example

Show the AI what "good" looks like by including an example of the style, format, or tone you want.

Example: "Write a product description in this style: [paste example]. Now write one for [your product]."

This is called few-shot prompting and dramatically improves consistency in style-matching tasks.

5. Say Who the Audience Is

The same content explained for a 10-year-old reads completely differently than for a PhD. Specifying the audience removes guesswork.

• "Explain this for someone with no technical background"
• "Write this for a C-suite executive who is time-pressed"
• "Assume the reader is a 16-year-old student"
• "Write for a professional developer who already understands REST APIs"

6. Ask for Options, Not Just One Answer

Asking for multiple variations lets you pick or combine the best elements rather than being stuck with one mediocre response.

"Give me 5 different subject line options for this email."

"Write 3 versions of this paragraph — one formal, one conversational, one humorous."

7. Iterate, Don't Start Over

After an initial response, refine it in the same conversation rather than starting a new prompt. The AI keeps context from everything in the current conversation.

• "Make it shorter — cut it by half"
• "The second paragraph is too technical — rewrite it in plain language"
• "Change the tone to be more direct and confident"
• "Add a practical example to point 3"

8. Tell It What Not To Do

Negative constraints are as powerful as positive instructions. If your responses keep including things you don't want, say so explicitly.

• "Do not use bullet points — write in prose paragraphs"
• "Do not mention competitors or make comparisons"
• "Do not start sentences with 'I'"
• "Avoid clichés like 'delve into' or 'it's worth noting'"