Export ChatGPT to Claude: Memory, Chats and Projects
November 11, 2026 · 10 min read
Export ChatGPT to Claude: Move Your Memory and Chats with Markdown
To export ChatGPT to Claude, you move context, not chat logs. Claude can't import ChatGPT conversation history, but it can import memory. Have ChatGPT write a structured Markdown summary of what it knows, paste the work-related part into Claude's memory import, and add the rest to a Claude Project.
What Can You Bring From ChatGPT to Claude?
Anthropic is direct about this. Its getting started article says it's not possible to import conversation history from another AI provider into Claude at this time. There's no official way to make your old ChatGPT threads show up as Claude chats.
What does carry over:
- Memory. Claude's memory import accepts a summary written by another assistant, and it's open to free accounts as well as paid ones.
- Reference material. Files you upload to a Claude Project stay available in every chat inside that project.
- Instructions. Your custom instructions become project instructions, written once and reused.
ChatGPT's own data export emails you a download link for a ZIP of every chat, including a conversations.json file. Keep it as your archive, even though Claude won't read it as history.
Should You Move Your Memory or Your Whole Chat Archive?
Most people need far less than they think. Memory covers who you are, how you like answers and what you're working on. That's the part you'd otherwise retype in every new chat, and it's what makes ChatGPT feel like it knows you.
The archive is different. It's reference material: a research thread, a long-running piece of writing, notes on a client. You don't need all of it, only the handful of topics you still return to.
We prefer a short context file plus three to five condensed topic files over dumping the raw export into Claude. A file that says "Decided on Postgres in March because of row-level security" beats 40 pages of the back-and-forth that led there. Claude reads less noise, and you can check every line before it goes in.
How to Export ChatGPT to Claude in Three Moves
Here's the whole migration at a glance. Each step gets its own section below.
- Ask ChatGPT to write your context. Run one prompt that makes it list what it knows about you under fixed headings.
- Clean it up as a Markdown file. Paste the result into a Markdown editor, cut anything private or stale, and save it as a .md file.
- Import and organise in Claude. Paste the work-focused sections into Claude's memory import, then add the remaining sections and your topic summaries to a Project.
This is how to export ChatGPT to Claude without a paid migration app or a browser extension. You only need the two assistants and a text editor.
The ChatGPT to Claude Prompt That Writes Your Context File
Start a new chat in ChatGPT and paste the prompt below. It asks for one code block, so the formatting survives when you copy it. Anthropic offers its own prompt on its import memory page, and that works too. This version adds fixed headings, which makes the result easier to split later.
List everything you know about me from your saved memories and our past
conversations. Output it as one Markdown code block with exactly these
headings: About me, Work and role, How I like answers, Ongoing projects,
Decisions already made, Terms and names. Use short bullet points. Write
ongoing projects as a task list with - [ ] for open items and - [x] for
finished ones. Leave out passwords, health details, financial details
and anything about other people that they wouldn't want shared.
Don't invent anything. If you're unsure about a point, mark it (unsure).
ChatGPT should reply with a single code block of Markdown. The code block matters: copying rendered text by hand often drops the # and - markers, while the block's copy button keeps them.
For topic files, open an old thread you still care about. Ask ChatGPT to summarise it as Markdown: the goal, what you decided, open questions, and any final text or code worth keeping. One file per topic keeps each one short.
Build Your Claude Context File in Markdown
Markdown fits this job because headings and bullets give a model clear boundaries. Claude can tell your preferences apart from your project list without guessing. Chat assistants already write their answers in Markdown, so both ChatGPT and Claude handle it well (see what Markdown is in AI).
A cleaned-up context file looks like this:
# Context for Claude
## About me
- Based in Leeds, UK
## Work and role
- Product designer at a 20-person SaaS company
## How I like answers
- Short paragraphs, then a bulleted summary
- UK spelling, no marketing tone
## Ongoing projects
- [ ] Redesign the onboarding flow (due March)
- [x] Pricing page copy
## Decisions already made
- We use Figma, not Sketch
## Terms and names
| Term | Meaning |
|---|---|
| OB | Onboarding |
| ICP | Ideal customer profile |
It renders as an H1 title and six H2 sections. Under them you get plain bullets, a task list with one open item and one finished, and a two-column table of terms. Paste ChatGPT's output into the editor below and edit it the same way.
The preview shows the task list as checkboxes and the glossary as a table, so you can spot a broken heading or a missing pipe before Claude sees it. The status bar's word count helps you keep the memory section short. When it reads well, use Download .md from the Download/Export menu. If the format is new to you, here's what a Markdown file is.
How Do I Import My ChatGPT Memory into Claude?
Claude's memory import takes pasted text. Per Anthropic's memory help article as of November 2026, the steps are:
- Open Settings > Your Claude (in the older Settings layout, Settings > Memory).
- Select Start import.
- Paste the sections about you, your work, how you like answers and your ongoing projects.
- Select Add to memory.
The new memory usually shows up within a short time, though some older Team and Enterprise setups take longer. To check or correct it, go back to Settings > Your Claude and select Manage next to Your memory, or ask Claude in a chat to change something.
Two caveats from the same article. Memory import is experimental and still in active development. Claude's memory also centres on work, so it may not keep imported personal details unrelated to work. Keep the hobbies and family notes out of this step; they belong in a project if anywhere.
Add Past Chats and Documents to a Claude Project
Projects hold the material memory doesn't. Per Anthropic's projects article, every plan has them, and the free plan allows up to five. Go to claude.ai/projects, select + New Project, and give it a name.
Then fill it in two places:
- Instructions. Select Set project instructions, paste your How I like answers section, and select Save instructions. This is where a custom GPT's instructions go too. For prompts you reuse, our AI prompt templates show the XML tag structure Anthropic recommends for Claude.
- Knowledge. Select the + button in the knowledge panel on the right and upload your topic summaries.
Anthropic's upload limits shape what's worth adding. Project files can be up to 30MB each, and there's no file count limit, but the total must fit Claude's context window. On paid plans, Claude also switches a project to retrieval (RAG) mode when its knowledge nears that limit.
One snag: .md isn't on Anthropic's list of supported types, which includes TXT, PDF, DOCX, HTML and JSON. If an upload is refused, rename a copy of the file to .txt. Markdown is plain text, so nothing is lost.
You can upload conversations.json itself if it's under 30MB, since JSON is supported. We don't recommend it. Years of chats can outgrow the context window, and most of the file is metadata and abandoned drafts. If you only want a small slice of structured data as a readable table, see our guide to converting JSON to Markdown.
Check What Carried Over After the Switch
Open a new chat inside your project and ask Claude what it knows about your work and current projects. Compare the answer with your context file, then fix gaps through memory settings or the project instructions.
Some things don't move at all. Anthropic documents a memory import only, so custom GPTs, generated images, voice chats and files stored in ChatGPT stay behind. Rebuild the ones you use as projects. Before you paste the context file, run it through the Markdown formatter if ChatGPT mixed * and - bullets or left stacks of blank lines.
The honest limitation: summaries lose detail. A condensed topic file keeps decisions and drops the reasoning, and occasionally ChatGPT summarises something wrongly. Keep the original export ZIP somewhere safe as your full record, and go back to it when a summary isn't enough.
Questions About Moving From ChatGPT to Claude
The fastest way to export ChatGPT to Claude is a single well-structured context file. It takes one prompt in ChatGPT, a few minutes of editing, then memory import plus a Project for the rest. You keep control of exactly what Claude learns about you. Paste ChatGPT's output into the editor, trim it until every line is still true, and download the .md file ready for Claude.