How to craft prompts that make AI work for you

post-thumb

Ever asked an AI chatbot for help and received something completely off-target? You’re not alone. The difference between AI writing that frustrates you and AI that feels like a productivity superpower often comes down to one thing: how you ask.

In today’s workplace, knowing how to communicate effectively with AI tools and large language models (LLMs) isn’t just a nice-to-have skill, it’s becoming essential. As Andrew Ng, founder of DeepLearning.AI and co-founder of Coursera explains, “When you use an instruction-tuned LLM, think of giving instructions to another person — say someone who’s smart, but doesn’t know the specifics of your task” (The New Stack, 2023).

The good news is that you don’t need a computer science degree to master prompt writing. This guide will help you craft better prompts for ChatGPT, Claude, and other AI assistants.

What’s in this article:

Why your AI prompts matter

Imagine asking a new colleague to “make a report about remote work.” Without more details, you’d likely get something that misses the mark completely. The same applies to generative AI—vague requests lead to disappointing results.

Let’s look at the difference in ChatGPT or Claude prompts:

Weak prompt: “Write content about remote work.”

Result: Generic, surface-level information about remote work that could apply to anyone, anywhere.

Strong prompt: “Write a 500-word blog post about the top challenges marketing teams face when working remotely, with practical solutions for each challenge. Include examples from small businesses. Use a conversational tone.”

Result: Targeted, useful content specifically addressing marketing teams’ remote challenges with actionable advice and relevant examples.

The difference is dramatic. With the stronger prompt, the AI clearly understands:

  • Exactly what topic to focus on (remote challenges for marketing teams)
  • How long the content should be (500 words)
  • What to include (practical solutions and examples from small businesses)
  • What tone to use (conversational)

The myth of brevity

Contrary to what you might think, shorter is almost never better when writing AI prompts. Many people worry they’re “bothering” the AI with lengthy requests. In reality, AI models like ChatGPT and Claude crave context. This means the more specific details you provide, the better results you’ll get.

As Isabella Fulford, a member of OpenAI’s technical staff, notes: “Clear writing doesn’t necessarily mean creating a short prompt, as in many cases longer prompts actually provide more clarity and context for the model, leading to more detailed and relevant outputs” (The New Stack, 2023).

Crafting prompts that deliver results

The anatomy of an effective prompt

A well-structured prompt helps AI systems understand exactly what you need and how to deliver it. Here’s how to structure your requests effectively:

1. Set the stage with context

Before diving into what you want, help the AI understand the bigger picture:

  • Who is the audience?
  • What’s the purpose of this content?
  • Where will it be used?

Example: “I need content for busy retail store managers who are considering implementing self-checkout technology but worry about customer satisfaction. This will be part of an email campaign.”

2. Be specific about what you want

Don’t make the AI guess what you’re looking for. State your needs clearly:

Instead of: “Give me some marketing ideas.”

Try: “Generate five Instagram post concepts that highlight our sustainable packaging for our eco-friendly soap brand. Each concept should include a suggested image description and caption that emphasizes how our packaging reduces plastic waste.”

3. Use a structured format for complex requests

Breaking your AI prompt into organized sections helps language models understand different components of your request, for example:

  • ROLE: Write as a personal branding consultant who specializes in LinkedIn strategy for tech entrepreneurs and has helped clients increase engagement by 300%
  • TASK: Create a comprehensive guide on crafting engaging LinkedIn posts that convert followers into leads
  • AUDIENCE: Millennial business owners (30-40) in SaaS and consulting who have LinkedIn profiles but struggle to get more than 5-10 likes per post
  • FORMAT: 800-word actionable guide with 5 specific post templates, 3 case studies, and a weekly content calendar
  • TONE: Conversational yet authoritative, using data-backed insights without corporate jargon
  • MUST INCLUDE: 3-2-1 content framework, optimal posting times (Tues/Wed 8-10am), storytelling formulas, carousel post structure, and a section on algorithm-killing mistakes like excessive hashtags and link-dropping

This structured approach works well for both ChatGPT and Claude, making your instructions clear and organized. For more customized help, check out prompt generators from Anthropic, OpenAI’s community favourite AI Prompt Generator GPT and Gemini.

4. Specify output format

Tell the AI exactly how you want information presented:

  • “Format this as a table comparing the three options”
  • “Present this as a numbered list of steps”
  • “Create this as a FAQ section with 5-7 questions and answers”

5. Show examples of what you like

Nothing communicates your expectations better than examples. Share samples that represent what you’re looking for:

“I’d like a product description similar to this example: [insert example]”

6. Set constraints to create better outputs

Setting boundaries helps focus the AI on what matters most:

  • Word count limits
  • Specific points to include or avoid
  • Formatting requirements
  • Tone guidelines

Example: “Create a customer email about our upcoming sale. Keep it under 200 words, avoid using exclamation points more than twice, don’t mention competitor brands, and maintain a friendly but professional tone.”

Building the feedback loop

Even the best prompts sometimes need refinement. The secret to getting exceptional results is treating AI interaction as a conversation, not a one-and-done request.

Iterate with specific feedback

When you get an AI response that’s not quite right, don’t just start over. Tell the AI what needs improving:

Instead of: “That’s not what I wanted. Let’s try again.”

Try: “This is good, but the tone is too formal for our audience. Could you rewrite it with a more casual, friendly voice while keeping the same information? Use contractions and simpler language.”

Build your prompt library

When you craft a prompt that works exceptionally well, save it. Create a collection of your “greatest hits” prompts that you can reuse and adapt for similar tasks in the future.

Consider creating templates for your common needs:

  • Meeting summary template
  • Customer email response template
  • Social media post template
  • Data analysis request template
  • Content creation template

Having a prompt library saves time and ensures consistent results across your AI interactions.

When AI hallucinates

Sometimes AI tools “hallucinate”, this means they generate information that sounds plausible but isn’t accurate. This happens when the AI has to fill in gaps in its knowledge or understanding of your request.

OpenAI’s training materials suggest a specific technique to address this: “The next tactic is to ask the model to check whether conditions are satisfied. So if the task makes assumptions that aren’t necessarily satisfied, then we can tell the model to check these assumptions first, and then if they’re not satisfied, indicate this and stop short of a full task completion attempt” (The New Stack, 2023).

To reduce hallucinations:

  • Provide specific information the AI might not know
  • Ask the AI to cite its reasoning or indicate uncertainty
  • Break complex requests into smaller, more manageable pieces
  • Review outputs critically, especially for factual claims
  • Request the AI to check assumptions before proceeding

Example prompt addition: “If you’re unsure about any specific statistics or facts, please indicate this rather than making up information.”

The 3-step prompt engineering cheat sheet

Remember these three principles for better AI results with large language models:

  1. Provide context: Help the AI understand the situation, audience, and purpose of your request.
  2. Be specific: Clear, detailed instructions yield better results than vague requests.
  3. Build on it: Use feedback to refine outputs until they meet your needs.

With these simple prompt engineering techniques, you’ll transform your interactions with ChatGPT, Claude, and other AI tools from frustrating to fantastic, no coding required.

Learn more about effective prompting

If you are interested in diving deeper into prompt engineering and AI writing techniques, check out these resources:

Ready to see how the right prompts can transform your work with AI? Autohive helps you collaborate with AI more effectively, turning complex tasks into simple solutions.

Try Autohive today and discover how the right prompts can make AI work for you.

Zheng Li

Written by

Zheng Li

Zheng Li is the VP of Marketing at Autohive. She is passionate about exceptional UX, product management and empowering every day people with AI technology.

You may also like