How to Move Your Memory From ChatGPT to Claude in 5 Steps

Everyone tells you to switch from ChatGPT to Claude. Almost nobody tells you how to actually move your memory over — the accumulated context of your writing style, projects, preferences, and recurring situations. Here's the five-step playbook, including the one hidden Claude setting that makes the whole thing work.

The Workflow in 5 Steps

Step 1 — Get ChatGPT to Describe Its Own Memory of You

Open a new ChatGPT conversation and paste a prompt that asks it to export everything it remembers about you — facts, preferences, projects, writing style, communication patterns. Save the result. This is ChatGPT's internal summary of who you are.

Step 2 — Export Your Full ChatGPT Conversation History

Inside ChatGPT:

1. Click your profile
2. Go to Settings → Data Controls
3. Click Export Data

ChatGPT will email you a zip file of all your conversations. Open it and extract the file called conversations.json — that's the raw, unabridged history of what you've talked about.

You now have two artifacts:

Artifact	What it is
Step 1 output	ChatGPT's self-described summary of you
conversations.json	Every conversation you've ever had, verbatim

Step 3 — Enable Memory in Claude (The Hidden Setting)

This is the step everyone misses and the whole chain breaks without it.

In Claude:

1. Open Settings
2. Go to Capabilities
3. Turn on the two buttons under "Memory"

Without this, Claude has no persistent memory surface to write into. You can paste 50MB of context at it and it evaporates at the end of the session.

Step 4 — Feed Claude the Combined Memory

Start a fresh conversation in Claude. Upload:

• The Step 1 summary (ChatGPT's description of you)
• The Step 2 conversations.json file

Include a prompt that tells Claude to analyze both, extract the durable facts about who you are and what you work on, and write them into its memory.

Claude will process both inputs and update its memory with the same model of you that ChatGPT had — but now stored in Claude's memory system.

Step 5 — Sanity Check

Start a fresh Claude conversation (no uploads, no prompts). Ask Claude what it knows about you, your projects, and your preferences. If Step 3 was enabled and Step 4 worked, Claude should surface your key context without any prompting.

Why This Matters

The switching cost on AI tools is almost entirely memory. A new model is a free swap. Losing your accumulated context is not. This workflow is the closest thing to an export/import pipeline between AI vendors — and it's possible because:

• ChatGPT gives you an official data export
• Claude has a memory setting (newly added)
• You can drive the conversion with a well-structured prompt

No API required. No third-party tooling. Just the two products plus a prompt.

The Bigger Picture

The fact that this even works is a partial answer to the "memory as moat" argument. Vendors lock you in, but they can't fully lock you in as long as:

• They offer a data export (GDPR / regulatory pressure)
• A competing product has an import path
• LLMs are capable enough to bridge one to the other

Expect this kind of migration pattern to become a standard workflow across AI tools — or to eventually be replaced by an open memory layer (the "Open Brain" pattern) that you own regardless of which model you use.

Key Takeaways

• Moving memory from ChatGPT to Claude is a 5-step workflow, not a migration tool — no one ships this natively
• Get ChatGPT's self-summary, then export the full conversations.json from Settings → Data Controls → Export Data
• The hidden step: turn on Memory in Claude → Settings → Capabilities before the import or nothing persists
• Feed both the summary and the JSON into Claude with a prompt and let it write to its own memory
• Switching AI tools costs accumulated context, not money — this workflow reclaims it

Resources

• ChatGPT Data Export — Official instructions for exporting your data
• Claude Memory Documentation — How Claude's memory feature works