Skip to content
OperatorNest

How to switch from ChatGPT to Claude without losing your context

I would move one working task to Claude before trying to move a whole ChatGPT history. Import memory, copy the instructions separately and check whether the new task still follows your preferences.

Chaitanya Karra · Member of Technical Staff · Published 27 September 2026, updated 3 October 2026 · 6 min read

On this page

What’s in a ChatGPT data export

I would save a ChatGPT data export before moving any work. The ZIP keeps your conversation history as raw JSON. To request one, go to Settings, then Data controls, then Export, and confirm on the follow-up screen. OpenAI’s help center says the export can take up to seven days to arrive by email, and the download link expires 24 hours after that, so you need to grab it promptly. It’s available on ChatGPT Free, Go, Plus, Pro and eligible Edu accounts; Business, Enterprise and Healthcare workspace users have to go through their workspace owner instead.

Sources checked 27 September 2026.

Use the ZIP as a searchable archive; deleted chats won’t return. OpenAI’s documentation doesn’t list saved memories, custom instructions, scheduled tasks or custom GPTs in it. Copy working context from the settings screens separately.

What ChatGPT memory and custom instructions store

ChatGPT keeps two separate layers of context, and they don’t move the same way. OpenAI’s memory help page describes saved memories as things you asked it to remember or that it saved as useful, and reference to chat history as a separate, looser layer that draws on past conversations and can change as ChatGPT updates what it thinks is useful. Both show up under Settings, then Personalization, then Memory, where you can review a memory summary and see the sources behind it, which can include past chats, saved memories, custom instructions, files in your Library, or connected apps like Gmail.

Custom instructions are a different setting: a text field where you tell ChatGPT what to know about you and how to reply, applied to every new conversation from then on. According to OpenAI’s custom instructions page, Free and Go accounts get up to 1,500 characters, while Plus, Pro, Enterprise, Business and Education accounts get up to 5,000, a limit OpenAI raised from 1,500 on 15 July 2026. Custom instructions shape how ChatGPT responds; they don’t give it knowledge it doesn’t otherwise have, which is worth knowing before you rely on them for facts rather than tone.

To carry either one to Claude, copy the text yourself: open the memory summary and the custom instructions field, and save what’s in them before you start the move.

Can you bring ChatGPT’s memory into Claude directly

Yes. Use Claude’s memory import and export tool. In Claude, go to Settings, then Memory, and select “Start import.” Claude gives you a prompt to paste into ChatGPT, asking it to list everything it has stored about you. Run that in ChatGPT, copy the output, paste it back into Claude, and click “Add to memory.” Claude extracts the text into individual memory entries you can review under “Manage edits.”

Anthropic warns that memory imports are “experimental and still in active development,” and “Claude may not always successfully incorporate imported memories.” What transfers well is work-related: preferences, instructions on tone and format, professional details, projects, tools and frameworks. Personal details unrelated to work may not stick, since Claude’s memory is designed around work context. The import doesn’t bring across your ChatGPT conversation history itself, only what ChatGPT reports knowing about you. Updated memory should appear shortly after import, though Anthropic notes it can take up to 24 hours for some accounts. The tool is available on Free, Pro, Max and Team plans, on the web and in Claude Desktop.

What Claude offers to rebuild the rest of your context

Memory import handles preferences and facts. For the deeper context ChatGPT projects and custom GPTs carried, Claude’s closest equivalent is Projects: self-contained workspaces with their own chat history and a knowledge base of uploaded documents that Claude uses across every chat inside that project. Projects are available on every plan, including free accounts, which get up to five.

For instructions that should apply everywhere, not just one project, Claude has a separate account-wide setting called Instructions for Claude (Settings, then Profile), which is the rough equivalent of ChatGPT’s custom instructions field. For the services ChatGPT connected to, Claude has its own Google Workspace connectors for Gmail, Calendar and Drive, available to all Claude and Claude Desktop users, added and authorized separately from anything you connected on the ChatGPT side. And for recurring work, Claude Cowork has its own scheduled tasks on Pro, Max, Team and Enterprise plans, built the same way you’d set up a scheduled task in ChatGPT: describe the job once, pick a cadence, and it runs.

Rebuild each separately. Memory import supplies context, without copying your ChatGPT setup.

What doesn’t transfer, no matter what you do

ChatGPT feature What happens when you switch Rebuild in Claude as
Saved memories and reference chat history Partial: work-related items import via Claude’s memory tool Memory (Settings, then Memory)
Custom instructions Doesn’t import; copy the text yourself Instructions for Claude (Settings, then Profile)
Full chat history Exports as raw JSON, not usable as live context Paste specific past chats into a project as needed
Custom GPTs No import path; OpenAI is retiring GPTs in favor of Plugins, migration targeted for 17 September 2026 and Enterprise retirement set for 11 December 2026 A Claude project with the same instructions and files, rebuilt manually
Connected apps (Gmail, Drive, Slack) Each connection and its permissions are ChatGPT-specific Reconnect and reauthorize per service in Claude’s connectors
Scheduled tasks OpenAI’s own help page confirms they’re “not currently included” in data exports Recreate the prompt and cadence in Claude Cowork’s scheduled tasks

A step-by-step migration checklist

  1. Request your ChatGPT export (Settings, then Data controls, then Export) so you have a personal archive, and download it inside the 24-hour window once it arrives.
  2. Open your ChatGPT memory summary (Settings, then Personalization, then Memory) and note anything important, since import can miss entries.
  3. Copy your custom instructions text from Settings, then Personalization, so you have it to paste into Claude’s Instructions for Claude field.
  4. Run Claude’s export prompt in ChatGPT. In Claude, go to Settings, then Memory, then Start import, and follow the prompt it gives you.
  5. Paste ChatGPT’s answer into Claude’s import screen and select “Add to memory,” then check the entries under “Manage edits” for anything that didn’t come through.
  6. Paste your custom instructions into Instructions for Claude, adjusting tone since the two fields aren’t scored the same way.
  7. Rebuild any custom GPT you rely on as a Claude project with the same files and instructions.
  8. Reconnect the apps you use, like Gmail or Drive, through Claude’s own connectors.
  9. Recreate any scheduled task in Claude Cowork with the same prompt and cadence.
  10. Check back after 24 hours and ask Claude what it learned about you, to confirm the memory import took.

Running ChatGPT and Claude side by side

For example, Sam keeps ChatGPT for quick drafts and uses Claude for longer-document research. Each handoff needs context: Sam must carry decisions from one conversation into the other.

Claude’s memory import narrows that gap for preferences and standing facts, since you can re-run it whenever ChatGPT’s memory picks up something new. It doesn’t close it for an active task. If Sam is mid-way through a competitor brief in ChatGPT and switches to Claude partway through, that brief’s research and drafts don’t come along, only what was saved to memory beforehand.

When this isn’t worth doing

If you use ChatGPT for a handful of one-off chats with no saved memory or custom GPTs, there’s nothing to migrate, and setting up Claude fresh is faster than exporting anything. The migration checklist above earns its time mainly if you’ve built up months of preferences, a custom GPT you rely on, or scheduled work you don’t want to write from scratch. Review imported entries under “Manage edits”, especially sensitive ones.

Moving between vendors means rebuilding context. Keeping it outside either vendor avoids that repeat setup. OperatorNest keeps memory you can read, edit and export yourself, and lets you send a given task to whichever model fits it, so a switch between models doesn’t cost you your context.

Common questions

Can I move my ChatGPT memory to Claude automatically?

Yes, with Claude's memory import tool. You generate a summary from ChatGPT using a prompt Claude provides, paste it into Claude's import screen, and Claude extracts it into individual memory entries, though Anthropic still calls the feature experimental and it can miss things.

Does exporting my ChatGPT data include my memory and custom instructions?

Not in a form you can reuse directly. The export ZIP contains your chat history as JSON, but OpenAI's own help center doesn't list memory or custom instructions as export contents, so you copy those from their settings screens separately.

What happens to my custom GPTs if I switch to Claude?

They don't transfer. There's no GPT-to-Claude import, and OpenAI itself is retiring custom GPTs in favor of Plugins, with a migration flow targeted for September 2026 and Enterprise retirement set for 11 December 2026, so a GPT's instructions and files need to be rebuilt as a Claude project either way.

Can I use ChatGPT and Claude at the same time on the same work?

Yes, and plenty of people do, running research or writing on whichever model handles it better and copying the result across. The friction is that neither one knows what the other did unless you tell it.

Is there a way to avoid re-doing this migration the next time I switch models?

Keeping your working context outside either vendor is the only way to sidestep it. OperatorNest keeps memory you can read and export and lets you assign a task to any model, so switching models doesn't mean starting your context over.

Hand off your first task tonight.

Tell us your email and what you'd hand off first. We'll send your access details and help you set up your operator.