How to move a Claude conversation into ChatGPT

Both of these live in a browser and neither will hand the other anything. Claude keeps no readable copy on your machine, so every option below is really a question of how much of the conversation survives the trip.

The four ways, worst to best

1. 01

   Ask Claude to summarise it, then paste the summary

   Fastest, and the one most people try first. It also throws away the thing that made the conversation worth moving: the wrong turns, the corrections, the constraint you mentioned once in the middle.

   A summary is Claude's account of the conversation, not the conversation. ChatGPT then works from a description of your work rather than your work.

2. 02

   Select all, copy, paste

   Keeps every word, which is the right instinct. The cost is that a browser conversation is a scrolling list of rendered messages, so you get the interface along with the text: the names, the button labels, the copy-code affordances. The paste arrives without any of the structure that told you who said what.

   It also stops working exactly when it matters. A conversation long enough to be worth moving is usually long enough that pasting it fills most of what the next model can hold.

3. 03

   Export your Claude data and find the conversation in the file

   Claude will email you an archive of everything you have ever said to it. That is a complete and faithful copy, and it is the slowest possible way to move one conversation: you wait for an email, download an archive of your entire history, then go looking for the one thread you wanted.

   Worth doing once as a backup. Not worth doing when you are mid-problem and want to keep going.

4. 04

   Hand it over in one keystroke

   Sidq reads Claude in your own browser, as you use it, already signed in. Press the shortcut, pick the conversation, and it writes a file you attach in ChatGPT. The whole exchange, word for word, never summarised.

   Attaching rather than pasting is the part that matters at length: the file goes to retrieval instead of into the context window of every following turn.

More questions: how Sidq reads each assistant · what it costs