Guide 5 min read

How to Transfer Your AI Memory to Claude

Switching AI assistants does not mean starting from scratch. Here is how to transfer everything ChatGPT, Copilot & Co. know about you – your preferences, writing style, workflows, and context – to Claude in about ten minutes.

Orcha Team

February 20, 2026

If you have been working with ChatGPT, Microsoft Copilot, or another AI assistant for months, you have left a lot of context there – writing style, preferred formats, recurring tasks, industry-specific knowledge. All of that is stored in saved memories and settings. And that is exactly why many people hesitate to switch: nobody wants to start from zero.

The good news: you can take that context with you. Anthropic offers a dedicated import function at claude.com/import-memory – but the default prompt there often is not enough to capture everything. A two-step approach gives you significantly better results.

Here is the full process – it works with ChatGPT, Copilot, and virtually any AI assistant that stores conversation history.

Why Claude is interesting for Finance & Business

Claude has evolved enormously in recent months – especially in areas relevant to finance teams. Long, structured documents such as contracts, invoices, or compliance texts are processed reliably by Claude, backed by an impressive context window. For analytical tasks – summarizing data, identifying patterns, drafting decision templates – Claude delivers results that are often more precise and better structured in practice.

On top of that, Anthropic places great emphasis on safety and responsible AI use. For companies working with sensitive financial data, that is not a nice-to-have but a requirement. Claude hallucinates less frequently, is more upfront when it is uncertain, and can be steered very precisely with clear instructions.

Step 1: Export your AI memory

Open a new chat with your current assistant – whether ChatGPT, Microsoft Copilot, or another tool – and enter the first prompt below. This exports stored memories, custom instructions, and any other saved information.

Then follow up with the second prompt in the same chat. This one goes deeper: it gets the assistant to summarize patterns from your entire conversation history – how you write, which tasks you regularly ask for, which tools you use, and how you want results formatted.

Both outputs together give you a complete picture of what your current assistant has learned about you.

Tip for Copilot users: The same prompts work with Microsoft Copilot as well. Copilot stores fewer explicit memories than ChatGPT, but the second prompt for history analysis still delivers valuable results.

Prompt 1: Export stored memories and settings

I'm switching to a new AI assistant and want to take my context with me. Please compile everything you know about me into a single, well-structured document. Include: 1. STORED MEMORIES List every memory entry you've saved about me — word for word as stored, without summarising or rephrasing. I need the exact raw entries. 2. CUSTOM INSTRUCTIONS / SYSTEM SETTINGS Reproduce any custom instructions or configuration I've set up, including: - Personal context ("What should the AI know about you?") - Response preferences ("How should the AI respond?") If any section is blank or not applicable, note that explicitly. 3. INFERRED PREFERENCES AND CONTEXT Beyond what's formally stored, list anything you've learned about me through our conversations: - My role, industry, and professional background - Software, platforms, and tools I work with - How I prefer information to be structured and formatted - My writing style and communication preferences - Subjects I ask about most often - Any specific conventions or standards I follow Be comprehensive — it's better to include too much than to leave something out.

Prompt 2: Derive behavioral patterns from conversation history

Now go a level deeper. Analyse our full conversation history — not just stored memories — and create a detailed profile covering: 1. COMMUNICATION STYLE How do I write and communicate? Describe my typical tone, sentence structure, vocabulary level, and any distinctive habits or patterns you've noticed. 2. MOST COMMON TASKS What kinds of requests do I bring to you most often? Rank them roughly by how frequently they come up. 3. ONGOING PROJECTS AND WORKFLOWS What recurring projects, processes, or workflows have appeared across our conversations? Include relevant context about their current state or my typical approach. 4. VIEWS AND PREFERENCES What opinions, preferences, or strong positions have I expressed that would help a new assistant understand my thinking? 5. CORRECTIONS AND BOUNDARIES Is there anything I've consistently pushed back on, corrected, or told you not to do? List any patterns of "don't do this" or "always do it this way." Use clear section headers throughout. This output will be imported into another AI's memory, so write it as a structured reference document — not as a conversational reply.

Step 2: Manually optimize before you import

This step is often skipped. But it is crucial to ensure you only transfer the relevant information. AI assistants do not only store useful context.

Copy everything your current assistant produced into a document of your choice – Google Docs, Notion, Apple Notes, etc. Then go through it once and remove what is no longer relevant: completed projects, one-off questions you will never ask again, and anything too personal to entrust to an AI.

What should stay:

Your communication style and preferred tone
Your professional role and industry
Ongoing projects and recurring workflows
How answers should be formatted and structured
Key preferences you do not want to explain every time

What you end up with is essentially your portable AI profile – a document you can use not just with Claude, but with any assistant. Create it once, use it everywhere.

Step 3: Import into Claude

Go to claude.com/import-memory and click "Start importing to Claude". This takes you straight to the memory settings.

Paste your cleaned-up context into the text box and submit it. Claude will process the information into its own memory system. It can take up to 24 hours for everything to be fully integrated.

Quick test: start a new chat and ask "What do you know about me?" – Claude should be able to reflect back the key details.

Note

Claude's memory feature requires a paid plan – Pro from $20/month. With the free version you can still use your profile: simply paste it at the beginning of a new chat. Claude will take it into account for that conversation but will not store it permanently for future sessions.

Tips for the best results

The export is thin? Then the memory feature was probably not enabled in your previous tool – or you have not been using it long enough. No problem: write your AI profile yourself. Note down your role, writing style, typical tasks, and formatting preferences. A hand-written profile is often even more useful than an auto-generated one.

You use Custom GPTs? Their system instructions are not included in the standard export. The equivalent in Claude is called "Projects": copy your GPT's instructions into a Claude project's system prompt and you have a comparable setup.

You want to be really thorough? In ChatGPT you can download your complete chat history via Settings → Data Controls → Export Data. OpenAI will send you a ZIP file by email. For most people this is overkill, but if you want to run a deep analysis of months or years of conversations, the option is there.

Conclusion

The entire process takes about ten to twenty minutes. The result: your new AI assistant knows your style, your way of working, and your preferences from the very start.

Want more guides like this?

Subscribe to our newsletter for new resources delivered to your inbox.