Claude has a different character from GPT models, and prompting it effectively requires understanding that. Here are eight techniques I have found genuinely useful after heavy use.
1. Give context about who you are and what you are trying to do
Claude responds better when it understands the situation. Not just “write a cover letter” but “I am a software engineer with 5 years of backend experience applying for a senior role at a fintech startup. Here is the job description.”
2. Ask for the format you want explicitly
If you want bullet points, say bullet points. If you want a numbered list, say so. If you want a flowing essay without headers, say that. Claude defaults to whatever seems reasonable but it does what you specify.
3. Use the word “only” strategically
“Give me only the code, no explanation” or “List only the drawbacks, I already know the benefits.” Claude is thorough by default - “only” focuses it.
4. Roleplay with a purpose
“Act as a skeptical investor reviewing this business plan” works better than “critique this business plan” because it gives Claude a clear perspective and attitude to adopt.
5. Ask for multiple versions
“Give me 3 different versions: one formal, one casual, one punchy.” Claude is good at this and the contrast often reveals what you actually wanted.
6. Use “think step by step” for complex reasoning
This is well-known but often skipped. For math, logic, or multi-step reasoning, explicitly asking Claude to think through steps dramatically improves accuracy.
7. Tell it what NOT to do
“Do not use bullet points. Do not use headers. Do not hedge with phrases like ‘it depends’.” Negative constraints are often more useful than positive ones.
8. Ask Claude to critique its own output
After getting a response: “What are the weaknesses in what you just wrote?” Claude is genuinely good at self-critique and often catches things it should have done better.
What are your go-to Claude prompting tricks? Share them.
Curated by Selendia AI ✍️
