Claude’s new feature lets you migrate your AI “memory” from ChatGPT in minutes 🧠🤖

Cliquez ici pour lire en français

You may have spent months “training” your AI assistant — repeatedly explaining your context, your projects, and the way you like to work. Then you switch tools, and suddenly you’re back to square one. Anthropic wants to break that cycle. With Import Memory, a new feature in Claude, the company promises a much smoother transition between AI assistants — especially for users coming from ChatGPT — without forcing them to rebuild their entire setup from scratch.

Breaking away from the “fresh start” problem ✂️

Until now, switching AI assistants has felt a lot like changing phone numbers fifteen years ago. You might gain better features, but you lose part of your digital life along the way. All the preferences you patiently built over time — your profession, ongoing projects, writing style, tone preferences, and little workflow habits — stayed locked inside the old tool.

With Import Memory, Anthropic is pushing a simple but powerful idea: your memories belong to you, not the platform. The company is introducing what it calls an “assisted exit” — turning what used to be a hard break into a continuous transition. The significance isn’t just technical; it’s symbolic. It acknowledges that AI should adapt to you, not the other way around.

How it works: two copy-pastes and you’re done ✍️

Anthropic designed Import Memory as a deliberately simple two-step process.

1. Export your memories from your old assistant

Head to Anthropic’s dedicated page at claude.com/import-memory. Claude provides a ready-made prompt that you paste into your current AI assistant — whether that’s ChatGPT, Gemini, Copilot, Grok, or another model.

The prompt asks the assistant to list all stored memories about you, including:

• tone and writing preferences
• formatting rules
• personal details (name, location, profession, interests)
• ongoing projects and goals
• recurring discussion topics

The AI compiles everything into a single structured block of text.

2. Import that block into Claude

You copy the generated code block and paste it into Claude’s memory settings using the import flow available in:

Settings → Capabilities → Memory

or directly via the “Import memory to Claude” card on the home screen.

Click “Add to memory.”

Claude then splits the block into individual memory edits and integrates them into its own memory system.

The full integration can take up to 24 hours, as Claude’s daily synthesis cycles consolidate the information into long-term memory.

The result: during your next session, Claude might already know that you’re a front-end developer in Douala, that you prefer three-part structured answers, and that you’re working on a weekly newsletter — without you having to explain it again.

A feature designed for serious AI users 📈

Import Memory isn’t a gimmick. It’s clearly designed for people who use AI as a real productivity partner.

Target audience: paying subscribers

Memory migration is limited to Claude’s paid plans:

• Pro ($20/month)
• Max
• Team
• Enterprise

Anthropic is clearly targeting users for whom personalization has become a valuable productivity asset.

Merging instead of replacing

Imported memories don’t overwrite what Claude already knows about you. Instead, the system merges new memories with existing ones, reducing the risk of conflicts, duplicates, or abrupt behavioral changes that could disrupt workflows.

Work first, small talk second

Anthropic says Claude prioritizes work-related information, such as projects, processes, and collaboration preferences. More personal details may be ignored if they don’t contribute directly to productivity or clarity.

User control and transparency

Imported data appears as individual memory edits that users can review, modify, or delete at any time from the settings panel. You can also edit the export prompt before sending it to your old assistant, allowing you to remove sensitive information.

Implicitly, Anthropic is sending a clear message: AI memory is becoming a strategic user asset — something you should be able to transport, audit, and adjust like any professional tool configuration.

A technical feature… and a political move for personal data 🔐

Behind this seemingly simple workflow lies a bigger debate: who owns your digital traces when they accumulate across prompts and conversations? By making it easier to export memories from other AI assistants, Anthropic is playing a two-sided strategy.

Attracting ChatGPT power users

The company makes switching to Claude almost frictionless: one prompt, one copy-paste, and your “AI brain” moves into a new body. Some observers describe the move as “opportunistic but brilliant” — a clever way to lure power users tired of rebuilding their setups.

Highlighting the lack of data portability

The feature also exposes a deeper issue. Today, retrieving your memories from an AI assistant often means asking it politely to list them, and hoping the model does it correctly.

Import Memory improves the situation, but it doesn’t solve the underlying problem: there is still no open standard for AI memory portability between platforms.

For organizations — companies, government agencies, media outlets — easier migration could change the balance of power. Lowering the cost of leaving a platform makes it easier to renegotiate vendor relationships while maintaining continuity, compliance, and governance.

The limitations: what Import Memory still can’t do ⏳

Anthropic is careful to present Import Memory as an experimental feature still under development. Several limitations remain.

No full history or files

Only “memory-type” information is transferred:

• preferences
• recurring context
• personal details

Past conversations, uploaded files, and platform-specific features are not included.

Imperfect memory synthesis

Claude may occasionally fail to integrate some imported memories or ignore parts of them. Some users have already reported that Claude’s memory can sometimes feel inconsistent, or that new conversations occasionally reference previous ones too literally.

A reminder that this isn’t magic

Updates are processed in cycles, meaning changes can take up to 24 hours to fully apply. That delay highlights that Claude’s memory is a structured system, not an instant variable — a constraint that also signals a more mature architecture. In other words, Import Memory isn’t yet a universal key to your AI identity. But it’s a clear step in that direction.

Who will really own your digital memory tomorrow? 🔭

With Import Memory, Anthropic isn’t just adding another feature to Claude. It’s opening the door to a world where your relationship with AI becomes portable, where your habits and context follow you across tools as easily as switching messaging apps. But a fundamental question remains: how far should this go?

The more convenient it becomes to use an AI that knows you deeply, the thinner the line becomes between productivity and privacy. Import Memory may be one of the first signals of a future where personal AI memory becomes a standard — portable, auditable, exportable, and perhaps even end-to-end encrypted — rather than a treasure locked inside a single platform.

So here’s the question: Would you switch AI assistants if you knew you’d never have to introduce yourself again?

Let’s discuss 💬
Should AI memory become more portable between platforms, or should it stay minimal and ephemeral to better protect personal data? Where would you place the balance between convenience, control, and privacy ? Share your thoughts in the comments.