How to move a conversation from ChatGPT to Claude
Memory import, a handoff summary, or the full transcript as a file: three ways to continue a ChatGPT conversation in Claude. Prompts included.
A common situation: you've spent an hour on a problem in ChatGPT and want Claude's take. Or the other way round — you hit a message limit mid-task. You don't want to start over, and pasting a long chat runs into the input size and clutters the context. Here's what you can carry over with built-in tools, and how to do it cleanly.
First, decide what you're moving
Two different things get mixed up here:
| What | Example | How it moves |
|---|---|---|
| Memory about you | "I write TypeScript, work in fintech, keep answers short" | Claude's built-in memory import |
| A specific conversation | A 60-message discussion of a service's architecture | A summary or a transcript file |
Method 1. Import memory into Claude
In March 2026, Anthropic added a tool to Claude that imports memory from other assistants (Tom's Guide, PCWorld). It works like this:
- In Claude, open the memory import tool and copy the ready-made prompt.
- Paste it into ChatGPT. ChatGPT produces a summary of what it knows about you.
- Copy that answer and paste it back into Claude.
As of October 2026, what moves is memory: your preferences, work context, preferred formats. It does not move conversations. If you need a specific chat, read on.
Method 2. A handoff summary: fast and compact
This is the best option most of the time. Ask ChatGPT to compress the conversation into a format another model can use. A prompt you can send at the end of the chat:
Write a summary of this conversation so I can continue it with another AI assistant.
Structure:
1. Goal: what we're trying to do (1–2 sentences).
2. Context: inputs, constraints, stack, important numbers.
3. Decisions already made, and why.
4. What we tried and dropped.
5. Open questions and the next step.
6. Important code or text — verbatim, in code blocks.
Be concise, no preamble. Don't add anything that wasn't in the conversation.
Then start a new Claude chat like this:
We're continuing work started with another assistant. The conversation summary is below.
First, briefly confirm how you understand the task and ask questions if anything is unclear.
Don't revisit options that were already dropped.
<summary>
…
</summary>
Why this beats a full copy:
- shorter: a 300–600 word summary instead of tens of thousands;
- cleaner: no typos, dead ends or outdated code versions;
- checkable: Claude restates the task first, so you see immediately if something got lost.
The downside: the model may drop a detail it considered unimportant. That's why the prompt asks for code verbatim.
Method 3. Move the whole conversation as a file
When every detail matters (legal text, a long debugging session, an interview), move the whole conversation. Pasting it into the input box is awkward; attaching a file works well.
- Save the ChatGPT chat as Markdown (see How to export a ChatGPT conversation to Markdown).
- Attach the file to a new Claude chat with the paperclip.
- Say what to do with it:
The file is my conversation with ChatGPT about a database migration. Read all of it.
Then: (1) list the decisions made, (2) flag where you disagree and why,
(3) suggest the next step.
If you're moving the conversation into a Claude project, you can add the file to project knowledge so every chat in the project can use it.
Comparing the methods
| Method | Speed | Fidelity | When to use |
|---|---|---|---|
| Memory import | 1–2 minutes | General context only | Switching to Claude for good |
| Summary | 2–3 minutes | Good, if you ask for verbatim code | Continue a task or get a second opinion |
| Transcript file | 3–5 minutes | Complete | Every detail matters |
The other way: from Claude to ChatGPT
All of this works in reverse. The summary prompt needs no changes: send it at the end of the Claude chat and paste the result into a new ChatGPT chat with the same opening line.
Two things to watch:
- Claude projects. If the conversation happened inside a project, some of the context lives in project knowledge and instructions, not in the transcript. A summary won't capture it. Move the important files separately — in ChatGPT you can put them in a project with the same instructions.
- Artifacts. Code and documents Claude created in the artifacts panel are best copied from there in full, rather than trusted to a summary.
Gotchas
- Input limits. Very long text in the input box may get cut off. A summary or a file avoids that.
- Images and attachments. Pictures and files from ChatGPT don't travel with the text. Attach them to the new chat again.
- The model argues with the summary. That's normal and even useful — it's why you wanted a second opinion. But if Claude starts re-solving settled points, remind it: "point 3 of the summary is final".
- Privacy. You're moving text yourself, between two of your own accounts. Check the conversation for passwords, tokens and personal data first.
What Chatsavo does today, and what's coming
Today Chatsavo helps with the first step — finding the conversation: full-text search across your ChatGPT and Claude chats with stemming, jumping straight to the matching message (filters and searching both sites at once are in Pro).
Export to Markdown with code, tables and a link to the original chat, ready to attach in Claude, is coming soon. A "Continue in another model" button that gathers the last messages or a summary, opens Claude and pastes the text into the input box is coming in the next updates. You'll still send the message yourself, after reviewing the text.
Chatsavo works on ChatGPT and Claude; Gemini, Grok and Alice are in testing; DeepSeek, Perplexity and GigaChat come later. If you work with both AIs regularly, keeping their chats in one panel with shared folders is simply easier. Try 7 days of Pro free, no card — installation.
Find any old chat in a second
Chatsavo adds folders and full-text search to ChatGPT and Claude. Free, and chat text stays in your browser.