Guide

Switch from ChatGPT to Claude without losing your context

Moving to Claude takes an afternoon. What you lose is the context: months of ChatGPT knowing your projects, your clients and how you like things done. Here is what carries over, what doesn’t, and how to set things up so the next switch costs nothing.

Steps checked against Anthropic’s help on October 21, 2026.

What carries over

What moves from ChatGPT to Claude
From ChatGPTMoves to Claude?How
Saved memoriesYes, mostlyClaude’s memory import (below)
Chat historyNoData export, then files in a Project, as reference only
Custom instructionsBy handPaste into Claude’s profile preferences
Projects and their filesBy handRecreate as Claude Projects, upload the files
Custom GPTsBy handCopy instructions and files into a Project
What you did outside ChatGPTNever was thereNeeds its own source (below)

Import your memory

Claude has an import for memories from other assistants, on Free, Pro, Max, Team and Enterprise, on the web and in Claude Desktop:

  1. In ChatGPT, ask it to list every memory and piece of context it has stored about you, in one code block, with dates, without summarizing. Anthropic’s help article has a ready-made prompt.
  2. Read the output and delete anything you don’t want to carry over.
  3. In Claude, open Settings › Memory and choose Start import. Paste, then Add to memory.
  4. Choose See what Claude learned about you and correct anything wrong.

Anthropic says the import is experimental and that Claude may not keep personal details unrelated to work. Expect to fix a few entries.

Old conversations

The import doesn’t bring chat history. If a few old conversations matter (a long planning thread, a contract you worked through), export your data from ChatGPT’s settings, find them in the export, and add them to a Claude Project as files. Claude can then read them in that Project’s chats.

Be selective. A whole export is mostly noise, and a Project with hundreds of old chats in it gives worse answers, not better ones.

Projects and custom GPTs

A Claude Project is the closest thing to both: instructions plus files, shared by every chat inside it. For each ChatGPT project or GPT you actually use, create a Project, paste its instructions, and upload its files. Leave the ones you haven’t opened in a month.

Why it comes out thin

After the import, Claude knows that you run a design studio and prefer short answers. It doesn’t know that the Acme redesign slipped to November, what you promised on Tuesday’s call, or which of the three logo options the client picked. ChatGPT mostly didn’t either: saved memory is a short summary of what came up in chats, and the import is a summary of that.

The context that makes an assistant useful at work is the recent, specific kind, and it was never in either assistant. It was in your calls, your inbox, your documents, your screen. That’s why every switch feels like starting over. More on this in why AI memory goes stale.

Keep context outside the assistant

The way to make the next switch free is to stop storing work context in one assistant. Keep it somewhere both can read:

  • Documents you maintain: a project brief per client, updated weekly. Works anywhere, takes discipline.
  • A memory connector: an MCP server that any assistant can read. Cachejay records your work on your Mac (screen text, calls, coding-agent sessions) and files it by project, and Claude reads it on the web, desktop and phone. ChatGPT can read the same memory on Business, Enterprise and Edu plans (Pro read-only; not Plus or Free).

With context outside the assistant, “what did we agree with Acme on Tuesday?” gets the same answer in Claude today and in whatever you use next year. Setup is on the Claude page.

Your first week

  1. Run the memory import and correct what Claude learned.
  2. Recreate the two or three Projects you use most.
  3. Paste your custom instructions into Claude’s preferences.
  4. Keep ChatGPT for a week for anything you can’t find, then cancel.
  5. Decide where your work context lives from now on, so you don’t do this again.

If you’re still pasting the same background into every chat, stop pasting context covers the habit, and give Claude memory of your work covers the rest of Claude’s options.

Questions

Can I move my ChatGPT memory to Claude?

Yes. Ask ChatGPT to list everything it remembers about you, then paste that into Claude under Settings › Memory › Start import. Anthropic calls the import experimental, and Claude focuses on work-related details, so check what it kept.

Does Claude import my ChatGPT chat history?

No. The memory import brings over what ChatGPT has saved about you, not your conversations. You can download your ChatGPT data and add old conversations to a Claude Project as files, but they stay documents, not memory.

Will I lose my custom GPTs?

They don’t transfer. Copy each GPT’s instructions and files into a Claude Project, which plays the same role: shared instructions and files for every chat inside it.

Does Cachejay import my ChatGPT memories?

No. Cachejay records your work as it happens (screen, calls, coding-agent sessions) and both assistants can read it. It doesn’t copy anything out of ChatGPT.

Stop explaining your day to ChatGPT

Install it, connect the apps you use, and get on with your work. Free for 14 days, no card.