I Switched from ChatGPT to Claude. Here's What I Wish I Knew.

I used ChatGPT daily for about eight months before I switched to Claude. The actual switch took maybe two minutes. But the first two weeks on Claude felt weirdly frustrating, and it had nothing to do with Claude being worse. It was the opposite, actually. Claude was great at the things I wanted it to be great at. The problem was that it didn't know me.

Every conversation felt like a first date. I kept having to re-explain things I'd already spent months teaching ChatGPT. My writing style, the frameworks I use, that I hate when AI responses start with "Great question!" ... all of it, gone.

That's what this post is about. Not a comparison of which tool is better (there are a thousand of those), but what the transition is actually like and how to make it less painful.

Why I switched

The short version: Claude's writing was closer to what I wanted. ChatGPT is great at a lot of things, but I was spending more and more time editing its outputs to not sound like a corporate press release. Claude just got the tone right more often, with less back-and-forth.

I also liked Claude's Projects feature for keeping context organized. And the longer context window made a noticeable difference for the kind of work I do, which involves a lot of "here are 15 pages of context, now help me with this specific thing."

Your reasons might be completely different. The point is, the "why" isn't the hard part. The "how" is.

The annoying part nobody talks about

When you switch AI tools, you lose two things:

Your conversation history. This one's obvious and honestly not that big a deal. Old conversations are reference material, not active context. You can still go back to ChatGPT and search them if you need to.

Your accumulated preferences. This is the one that hurts. Over eight months of daily use, ChatGPT had learned a lot about me. My job, my communication style, the tools I use, specific instructions I'd given it about formatting and tone. All of that knowledge is what makes an AI tool feel "tuned in" to you. And none of it transfers automatically when you switch.

It's like switching phones but none of your settings come with you. Same apps, but every single one needs to be reconfigured from scratch.

What actually transfers (and what doesn't)

Let's be clear about what's possible here. When you move from ChatGPT to Claude, you can transfer:

Your stored memories and preferences. Things like "I'm a frontend developer," "I prefer concise answers," "always use TypeScript examples." These are the things that shape how the AI responds to you, and they're the most valuable thing to bring over.

Your custom instructions. Anything you put in ChatGPT's "Custom Instructions" or "More about you" fields. These are essentially persistent system prompts, and they translate directly to Claude's memory system.

What you can't transfer: your full conversation history, any custom GPTs you've built, ChatGPT-specific integrations (like plugins or DALL-E generations), and your ChatGPT subscription. Those are platform-specific.

The good news is that the stuff you can transfer is the stuff that matters most for day-to-day use.

How I did it without starting over

Here's what I actually did, and what I'd recommend if you're making the same switch.

First, I exported everything ChatGPT knew about me. Not by manually copying things out of the Manage Memories screen (that only shows part of it), but by using a prompt that asks ChatGPT to dump its full memory, including things it inferred from conversations but never explicitly saved. We have a full guide with the exact prompt if you want to do this yourself.

Then I gave the export a quick scan and cleaned it up. I used the AI Memory Editor to review everything, removed a couple of things that were outdated (old job title, a project I'm not working on anymore), and fixed a few things that were just wrong. Took about a minute.

Finally, I pasted the cleaned-up export into Claude's memory import. Claude has a dedicated flow for this at claude.ai/import-memory, or you can find it in Settings. It took about 24 hours for Claude to fully process everything, but even right away it started picking up on my preferences.

5 things I wish I knew before switching

1. Export your ChatGPT memory before you get deep into Claude. I spent my first week on Claude manually re-teaching it things I'd already taught ChatGPT. If I'd exported first, I could have skipped all of that.
2. ChatGPT has two memory systems, and you want both. Saved Memories are the obvious ones. But ChatGPT also builds inferred preferences from your Chat History that don't show up in Manage Memories. The export prompt pulls both. I wrote more about this in What Does ChatGPT Actually Remember About You?
3. Claude's memory works differently, and that's fine. ChatGPT saves discrete memory entries. Claude synthesizes your context into a more holistic profile. The end result is similar, but don't expect an exact 1:1 mapping. Claude might phrase your preferences differently than ChatGPT did, but it'll still use them correctly.
4. You don't have to delete ChatGPT. I kept my ChatGPT account active for a month after switching, just in case. I still use it occasionally for specific things. Switching doesn't have to be all-or-nothing.
5. Give Claude a few days before judging. The first conversation felt off because Claude was still processing the import. By day three, it was responding like it had known me for months. The ramp-up time is real but short.

If you're on the fence about switching, the biggest thing I can tell you is: the memory problem is solved. That used to be the main reason not to switch. You'd lose months of built-up context and have to start over. That's not true anymore.

The ChatGPT to Claude migration guide walks through every step. Copy a prompt, paste the response, done. If you're thinking about making the move, just do the export first. You can always decide later.