Why Your ChatGPT Prompts Don't Work in Claude (And How to Fix It)
You switched to Claude expecting better output. Instead you got something that felt cautious, wordy, or just... off. Your prompts that worked perfectly in ChatGPT are getting mediocre results. So you assume Claude is worse.
It's not. You're using the wrong mental model.
The core difference
ChatGPT was trained to complete instructions. Give it a task, it executes. The mental model is: you're the boss, it's the worker.
Claude was trained differently. It responds best when you treat it like a highly capable collaborator — someone you're thinking alongside, not just delegating to. The mental model is: you're both working on the same problem.
This isn't a subtle difference. It changes everything about how you prompt.
The 3 mistakes ChatGPT users make in Claude
1. Over-instructing
ChatGPT rewards specific, detailed instructions. Claude finds them constraining. Over-specified prompts in Claude produce outputs that feel mechanical because you've left no room for it to bring judgment.
Instead of this (ChatGPT style)
"Write a 300-word email with 3 paragraphs. First paragraph introduces the problem. Second paragraph explains the solution. Third paragraph has a CTA. Use professional tone. No bullet points."
Try this (Claude style)
"I need to email a client who missed a deadline. The goal is to get a new commitment without damaging the relationship. What's the right approach?"
2. No context on why
ChatGPT doesn't need to know why you want something — it just does it. Claude uses context to make better decisions. The more it understands about your actual situation, the better the output.
Before your prompt, add 2-3 sentences of context: who you are, what you're trying to accomplish, and what "good" looks like. Claude uses this to calibrate everything — tone, depth, format, assumptions.
3. Treating it like a one-shot tool
ChatGPT is often used as a vending machine — one prompt, one output, done. Claude is better used as a thinking partner in an ongoing conversation. Start loose, refine together, get to something excellent over 3-4 turns instead of trying to nail it in one perfect prompt.
The Claude mental model in one sentence
Tell Claude what you're trying to accomplish and why. Give it the context it needs to make good decisions. Then get out of the way.
Once this clicks, Claude's outputs become noticeably better — more nuanced, more useful, more aligned with what you actually needed rather than what you literally asked for.
Want the full system?
The Claude Switcher's Playbook covers this in depth — 30 optimized prompts built specifically for Claude, a full translation guide for your existing ChatGPT workflows, and a Projects setup system for persistent context.
Get the full guide
The Claude Switcher's Playbook — $17
40+ prompts for switching from ChatGPT to Claude. Prompt translation guide, Claude Projects setup, and a 1-page cheat sheet.
Get it for $17 →