Anthropic has rolled out a streamlined way to bring your ChatGPT memories and personal preferences into Claude, cutting the pain of starting over with a new AI assistant. The feature guides you through exporting what another model knows about you and importing it to Claude in minutes.
The move lands as Claude’s popularity climbs, with its iOS app recently topping the Apple App Store’s free chart, and as some users publicly explore alternatives to ChatGPT. It is a classic play to reduce “switching costs” by preserving the personalization people rely on day to day.
For anyone who has spent months shaping a bot’s tone, workflows, and project context, this is more than convenience. It is continuity: your writing style, recurring tasks, and tool stack can come with you.
What the New Memory Import Feature Actually Does
Claude’s memory import pulls over structured snippets about you that another AI has learned: how you like replies formatted, topics you discuss often, personal and professional details you chose to share, and the software or languages you use. Anthropic says this works with major assistants, including ChatGPT, Google Gemini, and Microsoft Copilot.
Think of it as migrating a knowledge profile. Instead of retraining Claude with dozens of reminders, you hand it a curated digest so it can immediately mirror the context you built elsewhere.
How to Move Your AI Assistant Memories in Minutes
Start in Claude. Open Settings, go to Privacy, then Memory preferences, and choose Start import. Claude will generate a short request you can paste into the other AI to fetch your stored memories.
Next, visit the AI you are leaving. Paste Claude’s request into a new chat and submit. The message asks the assistant to list everything it has saved or inferred about you in a single, easily copied block. It typically covers your response preferences, personal details you provided, ongoing projects, and tools or frameworks you use, ideally with dates when available.
If you prefer to craft it yourself, a concise version might read: “Please export every memory and learned context about me, preserving my wording. Include instructions I gave you about tone and style, personal details I shared, recurring projects and goals, and tools or languages I use. Provide entries with dates if you have them, in a single block for easy copy.”
Copy the results. Before importing, review them in a text editor. Delete anything you no longer want stored or that is too sensitive for persistent memory. Many users discover old experiments or outdated roles that are better left behind.
Return to Claude and paste the cleaned list into the import field, then confirm Add to Memory. Claude will display a formatted summary of what it learned. You can verify by asking, “What do you know about me?” and checking for accuracy.
Privacy and Safety Considerations for Memory Import
Persistent memory is powerful, so treat it like any long-lived profile. Strip out home addresses, confidential client names, API keys, and anything regulated by your employer’s data policies. Importing only what you need keeps personalization without oversharing.
Anthropic lets you manage or purge memories at any time. In Claude’s Settings under Privacy, you can delete entries from chats or switch off “Generate memory from chat history.” You can also ask Claude to forget everything it has stored, and it should comply.
Why This Memory Import Matters for Power Users
Personalization is the stickiest part of any assistant. OpenAI has said ChatGPT surpassed 100 million weekly active users, and many of those users depend on consistent tone, toolchains, and project context. By offering a frictionless import, Anthropic turns curiosity about Claude into real trials, not just one-off prompts.
The timing is notable. Claude’s surge up the App Store and public campaigns urging users to reconsider default tools suggest a fluid market. Lowering migration friction is a proven growth lever in software, from email clients to note-taking apps—and now, AI assistants.
Troubleshooting and Edge Cases When Importing Memory
If the other AI returns little or nothing, check whether its memory feature is enabled. Some enterprise deployments disable memory by policy, and some regions or accounts may not have it activated. You can still export context manually by asking the model to summarize what it has learned about your preferences and projects from prior chats.
If formatting looks messy on import, remove bullets and emojis, keep one entry per line, and ensure each item reads as a standalone fact or preference. The cleaner the input, the better Claude will personalize from day one.
