How do I move my memory from ChatGPT to Claude (or Gemini)?

You've spent a year teaching one assistant how you work, and you'd like to try another without starting cold. Claude and Gemini both take an import from ChatGPT now — Gemini's, per Google, not yet in the EEA, Switzerland or the UK — and it's a five-minute copy-and-paste. Here are the vendors' own steps as they read on 2026-09-08, what genuinely comes across, and the part of your context that was never in there to move.

The short answer

Both moves are official (Gemini's outside the EEA, Switzerland and the UK, on a personal Google Account), and both are a copy-and-paste. You run a prompt inside ChatGPT that asks it to write out what it has stored about you, copy the reply, and paste it into a box in Claude's or Gemini's settings. Claude turns the paste into memory entries you can read and edit. Gemini does the same with a summary of your preferences, and separately lets you upload your entire chat history as a ZIP file.

What moves is memory text — the notes an assistant had written down about you from your conversations with it. What doesn't move, on either service, is anything that never entered a chat in the first place. That distinction decides whether the import gets you what you were hoping for, so it comes before the steps.

What you are actually moving

"Memory", as Anthropic's export prompt asks the old assistant to list it — and as Google's help article describes it, "preferences, remembered facts & details, and general contextual information (sometimes called 'memory')" — is a set of entries the assistant derived from what reached it: how you like things written, what you're working on, who you are. It's a by-product of the conversation, not the conversation. So there are three layers to keep apart:

• The memory entries. Short, derived statements about you. This is what both import flows are built to carry.
• The chats themselves. The full transcripts. Gemini takes these as a ZIP upload. Anthropic's import article covers memory only and doesn't describe a chat-history import into Claude.
• The material you were working from. The client's email thread, the scope document, the notes from the steering call, the voice memo in the car. If it was never typed or pasted into a chat, uploaded to one, or carried in through an app you connected, it never reached the assistant — so it isn't in either layer above. That last point is our reading of the vendor pages rather than their wording, and we come back to it.

How to import memory from ChatGPT into Claude

The steps below are read from Anthropic's help article Import and export your memory from Claude on 2026-09-08 (it shows as updated this week). Availability, in its words: "Memory imports are available for Free, Pro, Max, and Team plans on the web and Claude Desktop" — so do this at a desk rather than on the phone. Anthropic's pages speak of "other AI providers" throughout and never name ChatGPT; the steps are the same whichever assistant you're leaving. Anthropic's 25 August 2026 announcement describes memory as on by default on Free, Pro and Max plans "across web, desktop, and mobile", so, by our reading of the two pages together, what you import at the desk is the memory Claude carries elsewhere.

1. Ask ChatGPT to write out its memory

Anthropic supplies the prompt and says it's "written specifically to help you get all of your context in one chat." It opens: "I'm moving to another service and need to export my data. List every memory you have stored about me, as well as any context you've learned about me from past conversations. Output everything in a single code block so I can easily copy it." It then asks the assistant to cover — and "preserve my words verbatim where possible" — instructions you've given about how to respond; personal details ("name, location, job, family, interests"); "projects, goals, and recurring topics"; the tools you use; "preferences and corrections"; and "any other stored context", with an instruction not to "summarize, group, or omit any entries." Paste it into a ChatGPT conversation and copy the block it returns.

Two things the article adds are worth doing. "You can customize this prompt to exclude any sensitive information before importing into Claude" — read the output before you paste it anywhere; it's a client list, in effect. And: "if your current service supports generating files, you can also ask it to export your memory as a Markdown file so you can save a copy for your records." Keep that file. It's the one version of your memory that belongs to you rather than to a vendor, and it's useful below.

2. Open the import in Claude

"Go to Settings > Memory and select 'Start import.'" The import screen shows the same prompt with a text box under it. If your Claude shows Memory under Settings > Capabilities instead, you're on what Anthropic calls the legacy memory experience; the article gives near-identical steps from that path.

3. Paste and add

Paste the text and click "Add to memory." Then, in Anthropic's words: "Claude will extract key information and store it as individual memory entries. You can review the entries in the memory panel." The article also describes a "See what Claude learned about you" button that opens a fresh chat with "I updated my memory. What did you learn about me?" — the quickest way to see what survived the trip.

What Anthropic says about the limits

• "Claude's memory is designed to focus on work-related topics to enhance its effectiveness as a collaborator. As a result, Claude may not retain imported personal details unrelated to work." Anything you want kept regardless can be added through the "Tell Claude what to change or remove" box in Settings → Memory, or by asking in chat.
• "Memory imports are experimental and still in active development, and at this stage, Claude may not always successfully incorporate imported memories."
• The reverse trip is documented too. You can open "View and edit your memory" to see it "exactly how Claude sees it" (the article points at Settings → Capabilities for this; Anthropic's 25 August 2026 post says everything Claude remembers is "in a list of files under Topics in Memory settings, where you can read, edit, or delete each one"), or ask Claude to "Write out your memories of me verbatim, exactly as they appear in your memory" and "save this exported memory as a backup or bring it to another AI service by copying and pasting it into a local file on your computer."

How to import memory from ChatGPT to Gemini

Google's page is Import your AI chat history and memories to Gemini at gemini.google, with the detail in its help article Import from other AI platforms to Gemini Apps; both read 2026-09-08. Check where you are first. The help article states: "Important: This feature is currently not available in: The EEA, Switzerland, or the United Kingdom." It also requires that you "be signed in with a personal Google Account. This feature isn't available if you sign in to a work, school, or supervised Google Account", and that you "be 18 or older"; the public page adds that "compatibility and availability vary" and that Gemini's personalization must be switched on. Google's launch post of 26 March 2026 puts the same limits in a footnote: "Business, Enterprise and U18 accounts not supported. Not yet available for EEA/UK/CH users." Where it is offered there are two routes, and if you're leaving ChatGPT for good it's worth taking both. Unlike Anthropic, Google names ChatGPT and Claude in its export steps.

Route one — the memory

• Ask your current assistant. "Copy our suggested prompt and paste it into your other AI app to generate a summary of your preferences and style." The public page doesn't reproduce the prompt; Google's launch post says to "head to your Settings and select the new import option. You'll get a suggested prompt to copy and paste into your current AI app", and the help article gives the path as Settings & help → Import memory to Gemini.
• Copy the response. "Copy the personalised summary that it creates for you."
• Paste into Gemini. "Go to your Gemini Import settings, paste the text and start chatting with an AI that already knows how you work." The help article's wording: paste the response "into the designated text field, and click Add memory", after which "Gemini Apps will create a new chat thread and integrate this content to 'remember' these details for future conversations." Custom instructions travel the same way: "Just paste your preferences into Gemini's memory settings and you're good to go."

Notice the difference in what each vendor asks ChatGPT for. Google's route asks for a summary of your preferences and style; Anthropic's prompt asks for every entry, verbatim. If you kept the text file from the Claude steps, nothing stops you pasting that fuller version into Gemini instead — our suggestion, not Google's.

Route two — the chats themselves

• Export your data. "Request a data export from your current AI provider (they will usually email you a ZIP file)." Google's FAQ adds that you can "usually find this option in your other app's settings under 'Data controls' or 'Privacy'". For ChatGPT specifically, the help article's steps are: click your username at the bottom left, then "Settings → Data controls", and "next to 'Export data', click Export → Confirm Export"; "a download link will be sent to the email associated with your account on the other AI platform."
• Upload to Gemini. "Navigate to the 'Chat history import' section in your Gemini settings and select your ZIP file." Google supports ".zip files for your chat history imports (up to 5 GB)", and the help article adds two limits: "You can upload up to 5 .zip files per day", and the import "may take up to a day to complete." Before you upload a client archive, the help article's data-handling note: imported and continued chats "are saved in your Activity" and "this data is used to improve our services (including training generative AI models)"; Google adds "You can manage or delete your activity anytime."
• Pick up where you left off. "Gemini securely processes and organises your past threads so that you can search, reference and continue building on them." Imported chats appear "under 'Chats', marked with a special imported chat icon". To remove one, open its More menu and choose Delete; to remove a whole batch, the help article says to "find the import entry and next to the entry, click Delete" — "all imported chats within that .zip file will be deleted" (the public page calls that control the bin icon). Re-uploading a ZIP "will add any new conversations and overwrite previously imported ones."

What comes across — and what doesn't

Every cell below is either the vendor's wording or "the page doesn't say". Where a row is our own conclusion rather than theirs, it's marked.

Sources, all read 2026-09-08: Anthropic, "Import and export your memory from Claude" (support.claude.com, article 12123587) and "Switch to Claude without starting over" (claude.com/import-memory); Anthropic's 25 August 2026 post "Claude's memory works everywhere, and you decide what's in it" (claude.com/blog); Google, "Import your AI chat history and memories to Gemini" (gemini.google/import-memory), "Import from other AI platforms to Gemini Apps" (support.google.com/gemini/answer/16868299), and the 26 March 2026 launch post "Make the switch: Bring your AI memories and chat history to Gemini" (blog.google). OpenAI's help center could not be read on that date, so ChatGPT's side is described only as the Claude and Gemini pages describe it. All three vendors change these screens often — open the settings page in each app before relying on any row.

The workarounds, and where each one stops

Three habits make the import go further. Each is worth adopting; each has a ceiling you should know before you rely on it.

• Re-run the import when it matters. Nothing on either page describes an ongoing link between the services, so if you're using two assistants side by side, redo the export-and-paste before a piece of work where the newer context counts. The ceiling: you're the courier. It happens only when you remember, and it carries only what the source assistant will write out on the day.
• Keep the export as your own context document. The text file from step one is a portable brief. Paste it at the top of a new conversation in any assistant, on any plan, and you start warm. The ceiling: it's a snapshot you now maintain by hand, and it's still built from what an assistant noticed you say — not from the documents the work actually runs on.
• Use each product's containers. A project or Gem per client at least stops you re-explaining the basics inside that product. The ceiling: a container is scoped to its vendor, so it doesn't travel — and it holds what you put in it, which is the same limit again.

We work through these in more depth, with the vendor pages for each, in why your assistant forgets when you switch between ChatGPT and Claude.

Why your assistant still doesn't know your clients after the import

This is the moment that catches people. The import ran, Claude lists a tidy set of entries about your role and how you like your drafts — and the first real question, "where do things stand with the Acme renewal?", comes back thin.

The reason, in our reading, is structural rather than a fault in the import. The vendor pages establish that an import carries what the old assistant writes out; they say nothing about where ChatGPT's memory came from, so this next step is ours: that memory was derived from what reached ChatGPT — what you typed, pasted, uploaded, or carried in through an app you connected. Think about where the Acme renewal actually lives. The thread the client has been replying to all week, three messages deeper than the last time you read it properly. The revised scope attached to Tuesday's email. What their finance lead committed to on the last call — a line in a transcript. The voice memo you dictated in the car park because you knew it wouldn't survive the drive. Unless each of those was typed or pasted into a chat, uploaded to one, or reached it through a connected app, it was never in ChatGPT's memory, so it wasn't in the export, so Claude and Gemini had nothing to receive. The import did its job. The material was never in the box.

Which is why, however carefully you migrate, the feeling of starting cold comes back the next time a chat tool changes — or the next time you simply prefer the other one for a week.

A record that doesn't need moving

The alternative is to stop keeping the record inside any chat tool at all, and build it from the source material instead. That's what Equerry is: a personal assistant for iPhone and iPad, who holds what you send it and hands it back with the original attached — whichever chat tool you happen to be using this month.

• Built from the material, not from a chat about it. CC or forward the client thread to your own Equerry address (something like alex@mail.getequerry.app) and every later reply that keeps your address on copy lands too. Share the scope PDF or a screenshot from the share sheet, dictate the voice memo after the call, type the note. Nothing gets in unless you send it — you never hand over your inbox.
• Ask later, in plain language. "What did Acme's finance lead commit to on the scope change?" pulls across the thread and the attachment and your note from the call at once, because they sit in one archive rather than in three vendors' memory panels.
• Answers show their source. When an answer draws on your notes, emails or the web, it shows where it came from — not a derived line about it, the thing itself, one tap away.
• See and delete everything it kept. Every remembered fact is listed in one place and removable on tap, and every note you saved shows what Equerry took from it. Nothing to export to find out what it knows.
• Briefed before the call. Ahead of the Acme meeting, Equerry can brief you from what you've shared — including the reply that came in while you were double-booked.

Move your ChatGPT memory to Claude or Gemini by all means; the steps above take five minutes and the preferences are worth carrying. But the record that decides how Friday's call goes was never a chat memory, and it shouldn't have to be migrated every time your taste in assistants changes. Keep it somewhere that isn't inside any of them. See how an assistant that remembers across your tools works, or the side-by-side of ChatGPT memory vs Claude memory if you're still choosing.

Last updated 2026-09-08.

All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.