How to Move Your Context from ChatGPT to Claude (and Every Other Tool)
Extract, clean, and import your ChatGPT memories - and why portable context beats migration
You've been using ChatGPT for months. It knows your role, your projects, how you like to communicate. You want to try Claude for something - or you're switching entirely. And you realize: none of that context comes with you.
This guide covers how to move what you can, what you can't move, and why the better solution is context that never gets stuck in the first place.
What actually lives in ChatGPT that you want to move
Before migrating anything, it helps to know what you're working with. ChatGPT stores three distinct types of context:
Saved Memories - facts ChatGPT has accumulated about you from conversations. Your role, your projects, communication preferences, recurring topics. Managed under Settings → Personalization → Manage Memories.
Custom Instructions - what you've manually written in the two fields under Settings → Personalization → Custom Instructions. These are explicit, not inferred.
Conversation history - the raw transcripts of every session. Exportable as a JSON file, but not directly useful for context migration - it's hundreds of thousands of words of back-and-forth, not a structured profile.
One thing most migration guides get wrong: the official ChatGPT data export doesn't include your saved memories. Those are stored separately and don't come with the export file. You get conversation text, not the relationship.
Step 1: Extract your ChatGPT memories
Open a new ChatGPT conversation and paste this prompt:
I'm moving to a different AI tool and need to export everything you know about me. Please provide ALL of the following in a single structured response:
1. STORED MEMORIES
List every memory you have stored about me, verbatim as it appears. Don't summarize or paraphrase.
2. CUSTOM INSTRUCTIONS
Reproduce my full custom instructions exactly as written - both fields.
3. RECURRING PATTERNS
Based on our conversation history, list:
- Topics I return to most often
- My role and professional context
- Tools and workflows I use
- How I prefer responses structured
Copy the full output. This is your raw material.
Step 2: Extract your Custom Instructions separately
Go to Settings → Personalization → Custom Instructions and copy both fields manually. Don't rely on ChatGPT's recollection - copy the source directly.
Step 3: Clean and structure what you have
The output from Step 1 will be messy - a mix of accurate facts, outdated details, and things that were never true to begin with. Before importing anywhere, clean it up:
- Remove anything that's no longer accurate (old role, finished projects, tools you don't use)
- Organize into sections: who you are, how you work, what you're working on now
- Cut anything that's conversational noise rather than useful context
This is the step most guides skip, and it's why pasted memory often produces worse results than expected - garbage in, garbage out.
Step 4: Import into Claude
Option A: Claude's memory import (if available on your plan) Start a new Claude conversation and paste: "This is my context from another AI assistant. Please add this to your memory." Claude will integrate it into its memory system.
Option B: Claude Project instructions Create a new Project in Claude, go to Project Instructions, and paste your cleaned context there. This gives you more control and works on all plans.
Option C: Paste directly into a conversation For a one-off session, paste at the start of any conversation. Less persistent but works immediately.
What you can't fully migrate
Even after following all of the above, three things don't transfer cleanly:
Accuracy drift. ChatGPT's memories are inferred from conversations - which means they reflect what you talked about, not necessarily what's true about you now. If you mentioned a project six months ago, it's probably in there. If that project is finished, ChatGPT may not know.
Completeness. Memories only capture what came up in conversation. Things you've never explicitly discussed - your actual expertise, your full professional background, your current priorities - aren't there.
Portability. You've now moved your context from one platform silo (ChatGPT) to another (Claude). The next tool you want to try starts from zero again.
The better approach: context that doesn't live on any platform
The migration problem exists because your context is stored inside the tool that built it. Move tools, lose context. Every new tool is a clean slate.
The alternative is context that lives independently - extracted from your actual sources (LinkedIn, Notion, email, GitHub), structured into a portable format, and served to any tool via MCP. Not platform memory. Not a pasted document. A structured vault you own.
The practical difference: instead of extracting ChatGPT's imperfect recollection of you, you start from what's actually true - your real background, your actual current projects, your genuine expertise. And it loads automatically into every tool you use, not just Claude.
→ How personal context works: What Is Personal Context for AI?
→ How to build it from your actual sources: How to Build Personal Context for AI
→ Set up your context vault with Unabyss →
