The quality of an AI answer depends on the quality of the request. A vague prompt gets a vague answer. These 7 techniques work in ChatGPT, Claude, Gemini and most other assistants.
1. Give context first
Tell the AI who you are, who the output is for and why you need it.
Weak: Write a cold email.
Better: I run a small B2B software company. Write a cold email to IT managers at UK firms with 50 to 200 staff who may need better password management.
2. Say exactly what format you want
Ask for a table with 3 columns, 5 bullet points under 15 words each, or a 150-word summary. The AI will follow a clear format much more closely than a vague one.
3. Show an example
Paste an example of the style you like and ask the AI to match its tone and structure. One good example often works better than a long description.
4. Break big tasks into steps
Ask for an outline first, agree on it, then write one section at a time. You catch problems early and keep control of the result.
5. Ask it to reason before answering
For analysis or decisions, add “work through this step by step before giving your final answer”. It usually leads to more careful answers.
6. Set limits
Tell the AI what to avoid: “no buzzwords”, “do not invent statistics”, “under 100 words”. Limits stop the most common problems before they appear.
7. Improve the answer instead of starting over
If the first answer is close, give short feedback: “shorter”, “more casual”, “make point 2 more specific”. Treat it as an editing conversation.
A prompt template you can reuse
Context: [who you are and the situation]
Task: [what you need]
Audience: [who will read it]
Format: [length and structure]
Limits: [what to avoid]
Example: [optional]
Example: prompt for a sales email
Context: I sell a password manager to small businesses in the UK.
Task: Write the first line of a cold email to the prospect below, based only on the facts given.
Facts: [company name, recent news, job posting]
Format: 1 sentence, under 25 words.
Limits: no flattery, no claims that are not in the facts.
Prompts like this are the core of AI-assisted outreach. See how to use them at scale in our guide to personalising cold emails with AI.
Always check the output
AI assistants can state wrong facts with confidence. Check names, numbers, dates and quotes before you publish or send anything. The same applies to AI agents, which we explain in what AI agents are. For tool choices, read how to choose AI tools for marketing.
