Frustrated with AI Responses?

12 Tips from Anthropic That Actually Work

May 14, 2025

In partnership with

Receive Honest News Today

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Anthropic recently released a free guide on prompt engineering for their AI assistant Claude. In our view, it’s a very good resource for anyone looking to improve their AI interactions and better understand the magic behind prompting.

The guide focuses on practical techniques which, hopefully, will help transform your vague instructions into precise, consistent results.

Who's Claude, anyway?

For those new to the scene, Claude is a versatile AI assistant created by Anthropic, a research-focused company founded by former OpenAI leaders. Their mission centers on building AI that's helpful, honest, and safe. In other words: not simply powerful.

Claude comes in three main flavors:

  • Haiku: The speedy lightweight option for simple tasks

  • Sonnet: The balanced mid-range workhorse for most needs

  • Opus: The heavy-duty flagship for complex reasoning and problems

So…what makes Claude stand out from other large language models? Beyond the usual chatbot capabilities, it excels in several key areas:

  1. Safety and reliability: fewer mistakes and unwanted outputs

  2. Advanced reasoning: handles multi-step problems with clear explanations

  3. Long-context understanding: can process tens of thousands of words at once

  4. Multimodal capabilities: analyzes images alongside text

  5. Developer-friendly tools: straightforward APIs and versioning

Practical Prompt Engineering Tips

OK, folks, so let’s get to the good stuff…how to actually get better results from Claude (and, truth be told, any large language model):

Have Claude write prompts for you: Instead of struggling to craft the perfect instruction, simply ask: "Help me write a prompt that turns meeting notes into action items." You'll get a ready-to-use template you can adjust as needed.

Build reusable templates: Create standard formats for common tasks. Need frequent summaries? Save a template like "Summarize this report in three bullet points" and drop in new content whenever needed.

Use Claude as your prompt editor: When your instructions feel unclear, ask Claude to improve them. It's like having an editor for your AI instructions (meta, right?).

Be specific with your requests: Swap vague queries like "Tell me about health policy" for targeted ones: "Provide three advantages and two challenges of New Hampshire's telemedicine regulations."

Show examples of what you want: If you're looking for product descriptions, include a couple samples first, then ask for something similar. This technique (called multi-shot prompting) gives Claude a clear pattern to follow.

Ask for step-by-step reasoning: Instead of "What's the best route for snowplow scheduling?" try "List the factors you'd consider, then recommend an optimized route." This encourages Claude to break down its thinking.

Control structure with markup tags: Wrap sections in custom tags like <Summary>...</Summary> to organize Claude's output exactly how you need it.

Assign expert roles: Start with "You are a digital city planner" before asking about urban development, and Claude will respond from that specialized perspective.

Seed the response: Give Claude a starting point, like "Draft the first two sentences of a press release about our new recycling program," and it will continue in the same style.

Chain multiple prompts together: Break complex tasks into stages rather than trying to accomplish everything at once.

Handle large documents efficiently: When working with lengthy texts, give Claude a brief outline or upload in organized chunks rather than dumping everything at once.

Refine through follow-up questions: After the initial answer, ask Claude to elaborate on specific points or explain underlying assumptions.

I can pretty much guarantee these tips will improve your prompting game.