The one thing I wish I'd known before switching to Claude

By Ellis  ·  The Practical AI

When I switched from ChatGPT to Claude as my primary writing tool, I brought my ChatGPT prompting habits with me. That was a mistake — not because the habits were bad, but because Claude rewards a different approach.

Here's the thing nobody tells you when you make the switch.

Claude can hold much more context than you're used to giving it

With ChatGPT, I'd learned to keep prompts relatively lean. Give it the core instruction, maybe a sentence of context, and let it run. If the output was wrong I'd iterate — a follow-up prompt to adjust the tone, another to fix the structure. Short prompts, multiple rounds.

This works with ChatGPT because it's how the tool behaves best. Long prompts don't reliably produce better outputs; you're better off iterating.

Claude is different. It can genuinely hold a long, complex brief and follow it — not just the first few lines but the whole thing. Which means the correct approach is to invest more upfront. Give it the context, the constraints, the tone notes, the things you don't want. All of it, before it starts generating.

What this looks like in practice

A ChatGPT prompt for a blog post might look like this: 'Write a 1,000-word blog post about why most AI tool reviews are unreliable. First-person, informal tone.'

A Claude prompt for the same post looks more like this: 'Write a 1,000-word blog post about why most AI tool reviews are unreliable. First-person, informal but not casual — direct and slightly dry, like someone who has been burned by bad advice before and wants to be useful rather than entertaining. Avoid bullet points. The post should open with a specific scenario rather than a general claim. Don't use the word 'ultimately.' Target reader is an independent creator who is sceptical of AI hype but curious about what actually works.'

That second prompt is longer. It takes more time to write. And it produces a first draft that needs significantly less work. Over the course of a day's writing, front-loading the context saves time even though writing the prompt takes longer.

Why this matters for switching costs

If you try Claude with your existing ChatGPT prompts and find the outputs disappointing, this is probably why. You're using a prompting style optimised for a different tool.

The switch has a learning curve — not because Claude is harder to use, but because it rewards a different kind of effort. Once you adjust, it pays back. But the first week can feel like a step backwards if you don't know what's happening.

Give it two weeks with properly front-loaded prompts before making a judgment. The comparison post I published today has more on when Claude wins and when ChatGPT is the better choice.

Claude vs ChatGPT for content creators: the full 60-day comparison: https://thepracticalai.digitalpress.blog/claude-vs-chatgpt-content-creators/

Claude for Writers: the full review: https://thepracticalai.digitalpress.blog/claude-for-writers-review/

How to trial an AI tool properly before paying: https://thepracticalai.digitalpress.blog/how-to-trial-ai-tools/

— Ellis

This post contains affiliate links to Claude Pro. Full disclosure at /disclosure/