Move your ChatGPT memory to Claude, step by step

For anyone who wants to try Claude without rebuilding years of context about their work, tone and preferences.

By the end you will have exported what ChatGPT remembers, imported it into Claude, checked what Claude learned, and cleaned up what you do not want carried over.

The steps come from Anthropic's help article on importing and exporting memory.

What moves and what does not

• Moves: the memories and context the other assistant has stored about you, as text you paste into Claude.
• Does not move: your conversation history. Claude's help article states that importing past chats from another AI provider is not possible at this time.
• Claude's memory focuses on work-related topics, so personal details unrelated to work may not be kept.
• Memory import is experimental, so not every imported memory may be used.

Memory import is available on Free, Pro, Max, Team and Enterprise plans, on the web and in Claude Desktop.

Step 1: clean up first

Before you export, open ChatGPT's memory settings and delete anything you would not want in a new tool: health details, family names, anything private. Anthropic notes you can edit the export prompt to leave out sensitive information.

Step 2: get the export prompt

In Claude, go to Settings, then Your Claude, and choose Start import. Claude shows a prompt for the other assistant. If your layout is older, the path is Settings, then Capabilities or Memory.

Anthropic's prompt begins with "I'm moving to another service and need to export my data." It asks the assistant to list everything it has stored about you in one code block, with dates where available, including instructions, personal details, projects, tools and preferences, and to confirm whether the list is complete.

Use the prompt Claude shows you, since it may change. If you want to write your own, this version does the same job:

I am moving to another service. List everything you have stored in memory about me and everything you have learned about my preferences. Put it in one code block, one item per line, with dates if you have them. Group it under: instructions, work and projects, tools I use, preferences. At the end, tell me whether the list is complete.

Step 3: run it in ChatGPT

Paste the prompt into ChatGPT and copy the code block it returns.

Step 4: paste into Claude

Back in Claude, paste the text into the box and click Add to memory. Claude extracts the key details and stores them as separate memory entries.

Step 5: check what Claude learned

Click "See what Claude learned about you" to start a chat that asks Claude what it now knows. Read the answer and look for gaps or mistakes.

Timing: the newer experience says memory appears shortly after import. Users on the older layout may wait up to 24 hours.

Step 6: fix and extend

To change anything, go to Settings, Your Claude, and select Manage next to Your memory. Use the box that says to tell Claude what to change or remove, or ask in a chat.

You can also ask Claude to list your memories verbatim and save the text in a file as a backup.

After the move: a 10 minute check

1. Ask Claude: "What do you know about my business, my tone and my customers?"
2. Correct anything wrong in one message.
3. Test a real task, such as rewriting a product description in your voice.
4. Save a backup of your memory text.

What to do with old chats

Because history does not move, keep the valuable ones. Copy the useful prompts and answers into a notes file, or upload important documents to a Claude Project so the knowledge is permanent.

My honest take

The import saves the boring part: telling a new assistant who you are. It does not carry over your history, so treat it as a fresh start with a good briefing, and spend ten minutes checking the result.