How to Import ChatGPT and Gemini Memory into Claude

Why Move Your Memory to Claude?

Anthropic has introduced a feature that lets you bring your memory from ChatGPT and Gemini into Claude. This means your past conversations, preferences, and custom instructions can now live in one place. If you have been juggling multiple AI assistants, this could simplify your workflow significantly.

The feature is designed to make switching easier. You no longer need to start from scratch when you decide to use Claude as your primary assistant. Instead, you can carry over the context that matters most.

What Exactly Gets Imported?

The import covers the memory that ChatGPT and Gemini have built up about you. This includes things like your preferred response style, topics you frequently discuss, and any custom instructions you have set. It does not include full chat logs in most cases, but the memory summary is transferred.

This is useful if you have spent weeks or months training an assistant to understand your needs. Instead of repeating yourself, you can just point Claude to your existing memory.

Step-by-Step: Import Your Memory

Here is how to bring your memory from ChatGPT or Gemini into Claude. The process is straightforward, but you need to follow each step carefully.

• Step 1: Open Claude and go to the settings menu. Look for the option labelled "Import Memory" or "Transfer from Other Assistants." This is usually found under the account or preferences section.
• Step 2: Choose the source. You will see options for ChatGPT and Gemini. Select the one you want to import from. You cannot import from both at the same time, so pick the one that has the most relevant memory.
• Step 3: Authenticate the connection. Claude will ask you to log in to your ChatGPT or Gemini account. This is to verify that you own the memory you are trying to import. The process uses OAuth, so you do not have to share your password directly with Anthropic.
• Step 4: Review the memory summary. After authentication, Claude will show you a preview of what will be imported. This includes key preferences and instructions. Check this carefully to ensure nothing sensitive is included.
• Step 5: Confirm the import. If the preview looks good, click "Import." Claude will then transfer the memory and update its own memory settings with the new information.
• Step 6: Test it out. Ask Claude a question that relies on your past preferences. For example, if you had set a preference for concise answers in ChatGPT, see if Claude now responds in that style. This confirms the import worked.

What to Keep in Mind

The import is not a one-way street. After you bring memory into Claude, you can still edit or delete it. Go to the memory settings and you will see the imported items listed. You can remove anything you do not want to keep.

One limitation is that the feature works only with the memory summaries, not the full conversation history. So if you had a specific piece of information buried in a long chat, it might not come over unless it was part of the memory. For most users, this is not a problem because the memory is what Claude uses to personalise responses.

Privacy is another consideration. When you import memory, you are transferring data from one service to another. Anthropic says the transfer is encrypted, but you should still check what data is being moved. The preview step is there for a reason.

When to Use This Feature

This is the right tool if you are planning to switch fully to Claude from ChatGPT or Gemini. It saves time and reduces friction. It is also useful if you use multiple assistants but want Claude to have the same context as the others.

It is not the right tool if you only occasionally use different assistants. If you have not built up significant memory in either ChatGPT or Gemini, the import will give you little value. In that case, you are better off just starting fresh with Claude and letting it learn your preferences over time.

The feature is now available to all Claude users. Check your settings to see if the option is live for you. If it is not, it may be rolling out in phases.